Development of the methodology for the evaluation of a hydro-pumped storage power plant: Swiss case study

Energy Strategy Reviews 9 (2016) 8e17

Contents lists available at ScienceDirect

Energy Strategy Reviews journal homepage: www.ees.elsevier.com/esr

CASE STUDY

Development of the methodology for the evaluation of a hydro-pumped storage power plant: Swiss case study N.A. Iliadis*, E. Gnansounou Ecole Polytechnique Fe´de´ral de Lausanne (EPFL), ENAC, Bioenergy and Energy Planning (BPE), Ecublens 1015, VD, Switzerland

A R T I C L E

I N F O

A B S T R A C T

Article history: Received 20 October 2014 Received in revised form 19 October 2015 Accepted 21 October 2015 Available online xxx

During the last two decades, an evolving market structure is recognised starting from the spot market and its derivatives, resulting to several other markets such as the Intraday Market (IM), the Balancing Market (BM) and the Reserve Capacity Market (RCM). The participation in these markets has changed significantly the operation policy of the plants and, consequently, their value in the market. The current paper deals with the problem of the long-term valuation of a Hydroelectric Pump Storage (HPS) plant participating in the DAM and IM. For this purpose, an appropriate methodology has been developed while it is demonstrated on a real case study concerning the operation of a Swiss based HPS plant that participates in the DAM and IM of the German electricity market, for a horizon of thirty-five (35) years.

Keywords: Power plant operation policy Long term valuation Day ahead spot market Intraday spot market

Ó 2015 Elsevier Ltd. All rights reserved.

1. Introduction Since the launching of the first think-tanks and international forums on the impact of the power sector emissions on the environment, the generation mixes within countries started going through a significant change. There has been several studies analysing the effects of the fossil fuels on the environment and climate change speed [1] as well as the impact of cleaner generation technologies [2] such hydroelectric power [3], natural gas [4] and other technologies [5]. The latter studies are divided between these calculating the carbon emissions based on energy [6], and these having identified the mechanism of environmental effects and developed the relevant mathematical models [7, 8]. The common obvious result of these studies is the positive contribution of the renewable energy sources and low polluting technologies to the

* Corresponding author. E-mail address: [email protected] A. Iliadis). http://dx.doi.org/10.1016/j.esr.2015.10.001 2211-467X/Ó 2015 Elsevier Ltd. All rights reserved.

(N.

environmental health and climate change prevention. However, the important increase of renewable energy sources into the generation mix has raised a number of issues on the system reliability and marginal operating cost. The last two decades have been characterised by the shifting from a cost minimisation objective to the revenue maximisation of power plants. More specifically, under the context of deregulated electricity markets, generation companies have abandoned the traditional minimum cost operation optimisation schemes and have moved to an operational policy where daily decisions are taken determining the amount of power, along with the type of ancillary services, they are willing to provide in order to maximise their profits [9]. An electricity spot market may be organised as a sequence of different market mechanisms including Day-Ahead Market (DAM) and the Intraday Market (IM) (during the delivery day). Given this marketplace’s structure, every typical generation company tries to determine its optimal operation

strategy in order to maximise the respective revenues and thus calculate the value of the plant in the long term. To achieve this target, there is an increased need of an appropriate methodological framework to be adopted that simultaneously considers these multiple markets and addresses the problem of deciding in which market mechanisms the owner agent should participate and in which volumes will maximise its profits [10] while considering a long horizon. Furthermore, the increasingly growing penetration of wind and solar energy in all European countries has made forecast horizon shorter, emphasising thus, the importance of IM both for the demand and supply. Especially in countries like Germany, due to the high penetration levels of solar photovoltaics along with their generation profile particularities (production only in peak daytime), the difference between peak prices and off peak prices has been significantly reduced. Moreover, the highly possible lack of base load capacity, as a consequence of the nuclear plants decommissioning adopted by Switzerland and Germany, augmented further

N.A. Iliadis, E. Gnansounou / Energy Strategy Reviews 9 (2016) 8e17

the off-peak DAM prices reducing thus the peak to off-peak spread. Further to the structural changes above, the crisis that hit the world and Europe the last six (6) years, had as a result the important decrease of the demand and hence the continuous fall of the prices to a level as low as 50% of their 2009 value. All these changes clearly denote the difficulty of a high flexibility plant belonging to the so-called Energy Storage Technologies (EST), to collect sufficient revenues from participating only in the DAM and highlight the increased need to participate in the IM as well. The most representative and widespread known case of EST is the hydro pumpedstorage (HPS) power plants, which can adopt diversified bidding strategies acting in multiple electricity market segments and therefore being in a potential privileged position compared to the thermal power generating plants. Their advantage comes from the storage capability of HPS power stations allowing them to be involved in both generating and pumping, and buying and selling energy depending on the price profile and their flexibility. For this reason, an HPS utility can buy and sell electricity on the DAM as well as on the IM, by following an appropriate strategy that can result to revenues’ maximisation. The conventional way of estimating the future revenues of a HPS power plant considers simulating its activity in the DAM. Other revenues are then calculated as a fixed percentage on top of the latter amount. As pumping implies energy consumption, the economic efficiency of the operation can be achieved only if the electricity selling price is higher than the purchased price beyond a certain target, depending on the technical efficiency of the HPS plant. However, due to all the previously mentioned factors, the difference between peak and off-peak electricity prices is decreased, affecting thus the economics of HPS plants. All these aspects concur for the use of new instruments such as the IM, while the importance of potential revenues from the IM leads to the need of a more accurate consideration of them in the long term valuation of the HPS plants. The purpose of this paper is to describe the main aspects of a methodology that can be used for the long term valuation of a HPS power plant. That plant valuation consists in estimating the net expected present value of a plant taking into account investment and future revenues from operation in the various market that the plant participates in. A practical application of the developed methodology will be presented by addressing the main questions of calculating, developing and laying out the optimal simulation of the electricity generation policy of a Swiss HPS for its long term valuation (i.e. from 2015 until year 2050) through the optimal

allocation of power output being dealt to the successive market mechanisms of DAM and IM in Germany. The selection of Germany has been based upon the high liquidity achieved in the energy market being calculated as the ratio of the traded volume of wholesale dayahead power contracts and the annual gross inland electricity consumption in the respective region, among the possible markets that a Swiss operator can participate in. The solution to the assessed problem reveals the real value of an HPS plant in Switzerland, and consequently the assumed operation strategy, from a long-term perspective, required in order to lead to an expected revenue maximisation. Thompson et al. in Ref. [11] present an algorithm that can be used for the valuation and optimal operation of hydroelectric and thermal power plants operating in the liberalised electricity market. Appropriate equations are derived by using real options theory that incorporate an increased variety of spot prices models for representing realistically the main characteristics of spot prices markets (e.g. price spikes). The operational constraints of the involved power plants are taken into consideration and the developed models lead to non-linear partial-integrodifferential equations, when solved determine simultaneously the expected cash flows and the optimal strategy. The paper presents the obtained results for two case studies that concern a hydroelectric pumped-storage facility designed to take advantage of price spikes and a hypothetical thermal generator. Pedram and Moff in Ref. [12] focus on the valuation of energy storage technologies in electric power markets by assessing the relative arbitrage revenues that may occur. For this purpose, appropriate models have been developed for optimising the operation of a storage facility over a 24-h period while the respective problem has been framed as a linear program, a multi-stage stochastic program and a dynamic program. These discrete optimisation frameworks have been separately used in order to analyse two specific storage technologies. Botterud in Ref. [13] investigates the way that dynamic and stochastic optimisation can be used for improving the investment decisions concerning the power generation facilities in the competitive electric energy markets. An appropriate model has been developed for determining the optimal conditions for investing in a new power generation technology, when solving the respective problem using the dynamic and stochastic solution. Within this framework, a stochastic simulator is developed in order to realise the way that the investment strategy may differentiate through time, assuming that the investor changes the implementation of potential investment projects due to more

9

updated and accurate information being gradually available. A case study that concerns a new base-load gas power plant in the Norwegian power system has been assessed in order to demonstrate the revenues gained from the developed methodology. Weber and Woll in Ref. [14] present a methodology that has been developed for valuing the long-term operation of CHP portfolio by applying recursive stochastic optimisation. The main difference between the CHP power plants and the conventional power plants regarding their optimal operation in the liberalised market, concerns the additional operational constraint for the heat demand that needs to be satisfied. This constraint is appropriately modelled in the proposed method while a specific case study concerning a CHP system consisting of eight CHP power plants and two boilers with two main district heating grids is assessed. Siddiqi in Ref. [15] concentrates on longterm project evaluation and power portfolio management issues in a competitive electric energy market that include decisions about various aspects such as building a power plant, entering into contracts for buying or selling energy over the short or long-term, etc. The developed method is based on Options Pricing Theory and Decision Analysis and is applied for implementing the long-term power portfolio management of an existing power utility in the U.S. Deng in Ref. [16] addresses the problem of valuing electricity generation capacity as well as investing in new power generation facilities within the framework of the competitive electric energy market. An increased attention is given to the power price spikes that may occur and affect significantly the decision about investing in new generation capacity as well as the actual time to make this investment. Proven by the literature review above, it is justified that the present research is occupied with a subject, where scarce scientific work has been published. Therefore, an opportunity is offered to develop a new methodology covering the relative research gap between long-term valuation, various markets participation and HPS plants valuation, thus promoting new ways of addressing and solving this specific combination of problems. Furthermore, having taken into account the current needs of Swiss electricity utilities, this study is also of entrepreneurial importance for the local HPS operators. The obtained results regarding the Swiss case study are expected to be of increased importance, not only from a theoretical point of view but also on the actual implementation to HPS facilities. In recent years, increased research effort has been devoted in developing optimal policies for electricity utilities participating in

10

N.A. Iliadis, E. Gnansounou / Energy Strategy Reviews 9 (2016) 8e17

multiple competitive market mechanisms. However, the particular case of HPS power plants and the determination of their operation in the IM have not yet been adequately researched. An important characteristic of an HPS for the development of the operation policy in multiple markets is the storage capacity due to the reservoir and the “recharging” capacity due to the pumping system. Additionally, no work has been done in the development of a methodology that evaluates the operation of a plant in a longterm horizon considering both the DAM and IM. The proposed methodology of this research constitutes an approach focused exclusively on the HPS owner’s perspective seeking to evaluate a plant, presenting innovative characteristics and contributing to a new research field. Investigating and studying the existing literature on the short-term and mediumterm operation of a HPS plant participating simultaneously in multiple markets, it is seen that there are various approaches and related papers. They work on the development of methodologies focused on the combination of these two specific market mechanisms. Hence, focussing significantly more on the ancillary services markets such as the BM and the Automatic Generation Control (AGC). According to Deb in Refs. [17], the maximisation of revenues is the main goal of any operator acting in the environment of a multicommodity market, accessing all markets available. For a power plant, this fact is considered as actively participating in the ancillary services as well as the spot markets. In this paper is examined operation and revenues of an HPS plant and the impact the impact of the ancillary service markets. More specifically, a multi-product, multi-area structural model with an optimal power flow dispatch for real-time pricing is used in order to examine a typical HPS plant, simulating hourly in the DAM and in the ancillary services market. Furthermore, a typical hourly forecast is presented concerning the prices for various features such as energy, upward regulation, downward regulation, spinning, non-spinning and replacement reserves, and real-time energy. Eventually, the obtained prices are applied to the calculation of the optimal bidding strategies for the HPS plant. Pinto and Neves in Refs. [18], consider the penetration of RES in the system and focussing in the ancillary services, an optimisation model is developed in order to determine the optimal bidding strategies that allow the total revenue maximisation of a HPS plant acting in a competitive electricity market. The HPS plant is located in Portugal and participates both in the Iberian DAM and in the Portuguese ancillary services market. In order to calculate the optimal bidding strategies of the HPS station and quantify the impact of high wind

power penetration levels in the formed market prices, thus affecting the actual revenues of the plant, various scenarios assuming different levels of wind power facilities are examined. Following the scenarios’ results calculations, two major conclusions emerge: a) as the wind energy contribution in the DAM is expanded, the HPS plant revenue presents a relative increase, and b) the majority of the revenues earned in all scenarios are coming from the power plant’s participation in the spinning reserve market. Kanakasabapathy and Shanti in Ref. [19] consider the non-linearity arising from the technical specifications of a HPS plant, a tool is presented that allows a HPS power producer to optimally schedule the short-term operation of its asset in the DAM as well as in the ancillary services market. This tool corresponds to a model accounting for the non-linear three-dimensional relationship between the following three aspects: stored energy, reservoir head and power output. The optimal bidding strategies for a HPS power plant operating under the framework of a pool based competitive electricity market are investigated while a case study based on such a typical power plant is examined. The obtained results have proven the suitability of the developed tool for determining the optimal operational scheduling of the plan on a daily as well as a weekly basis. Loehndorf et al. in Ref. [20] model the optimal decision process of a storage operator who trades in a wholesale electricity market consisting of a DAM and a real-time market (BM). The decision-maker places multiple bids in the day-ahead auction e one hourly bid concerning the following day. When day-ahead prices are cleared, the operator can choose among using storage capacity or the BM to fulfil day-ahead commitments and close his positions. The operator maximises expected revenues by optimising the operation of the large-scale energy storage plant (e.g. HPS) while bounding the conditional value-at-risk (CVaR) to avoid overly speculative bidding decisions. The authors model a state-dependent, linear price-response function for the DAM in order to study the impact of large-scale storage behaviour. Wilde in Ref. [21] proposes a model considering the participation of a HPS plant in DAM and the RCM and attempts to assess the effect of RCM on the optimal bidding strategy and operation of the power plant by using an appropriately developed optimisation algorithm. HPS operators are able to provide spinning and non-spinning reserve capacity and might therefore have a strong incentive to adjust their optimal operation strategies and earnings potential. This consideration becomes increasingly relevant due to the rising feed-in of fluctuating wind energy in

many power systems, which influences future electricity spot prices and the need for reserve capacity. It is seen that when deciding the bidding strategy, the revenue opportunities in RCM need to be taken into account as they have an increased impact on the optimal plant operation. In addition to organised market, Lu et al. in Ref. [22] consider the Over the Counter (OTC) market and it is stated that an individual HPS plant owner buys and sells electricity either on the DAM and BM or with bilateral contracts. The optimal bidding strategy of an electricity utility with HPS assets in the DAM, constitutes the main issue of the paper. It is assumed that the revenues of an HPS plant comprise the selling energy during the power generation operational state and the participation in the nonsynchronous RCM during the non-generating operational state. Furthermore, due to the ability of the HPS plant to quickly reduce the respective pumping power when posed by the system operational constraints, such plant are considered to be appropriate for supplying synchronous reserve. As a result, the scheduling of the HPS operation is strongly incentivised towards the direction of power generation during the price peaks time periods, whereas the pumping procedures are performed during the lowest prices periods, in order to maximise the respective revenues gained. Loehndorf and Minner in Ref. [23] describe a model concerning the optimal bidding strategy of RES generation in the DAM assuming the existence of energy storage facilities, such as a HPS power plant. The model is developed as a continuous-state Markov decision process and a solution approach is presented based on Approximate Dynamic Programming (ADP) in Refs. [24], providing the possibility of solving large-scale problems. It is seen that better policies are obtained by using the described model compared to the respective ones obtained by applying a two-stage approach. The latter would be restricted to maximising expected immediate rewards. Olsson and Soeder in Ref. [25] present a short-term hydropower-scheduling model, expressed as a stochastic optimisation problem that is used in order to manage the tradeoff between energy and RCM. The relevant objective function comprises a. the sum of sales in the spot market and the regulating market and b. the value of saved water, while the respective constraints applied concern the hydrological balance as well as the balance in produced power and traded power in the market. This model has been developed because of the significant wind penetration levels achieved in the Nordic power system that result to increased requirements of regulating power in order to be used for

N.A. Iliadis, E. Gnansounou / Energy Strategy Reviews 9 (2016) 8e17

dealing with the respective wind power sudden variations. It is seen that hydroelectric power plants will probably be the facilities that will be called on to supply the relative requirements. Plazas et al. in Refs. [26], though not referring to an HPS plant, present a scientific work dealing with an interesting study about the revenue maximisation of a producer owning a thermal power plant which participates in the following spot markets: DAM, AGC market and BM. A price-taking behaviour is assumed for the producer in both the DAM and the AGC market whereas a potential price-making one is assumed regarding the highly volatile BM. The producer’s objective is to maximise its expected revenues from selling energy in DAM, BM and in AGC market. The problem emerging is manipulated by applying a multi-stage stochastic programming methodology in order to determine the optimal bidding strategies for the DAM. An equivalent deterministic problem is formulated which is efficiently solved using a commercial mixed integer linear programming (MILP) solver. Similarly, Corchero et al. in Ref. [27] propose a stochastic programming algorithm that can be used for determining the optimal bidding strategy of a power generation company operating in the DAM by taking into consideration the benefits and costs that occur by the participation in the subsequent markets as well, such as the IM and the RCM. The sequence of these three markets as well as their respective actual characteristics that concern the Iberian Electricity Market (MIBEL) are taken into account and a mixed linearly constrained minimisation problem with a convex quadratic objective function is formulated. The proposed model is validated through the assessment of an appropriate set of tests that concern a case study. Finally, Schuman in Ref. [28] presents an analytical model in order to establish an effective way for a wind power producer to deal with the prediction error, during a certain generation time, in the IM. The model is used for calculating the expected revenues that may be obtained by using the best predictions available in order for the producer to adjust appropriately the respective bid in the IM. The obtained value is compared with the respective one concerning a less active producer in the IM. The literature review shows a quite limited number of attempts to include two or more market mechanisms under the perspective of simulate the operation that could be used in the long-term study especially for a HPS power plant operator. As shown, many of these papers consider as second market mechanism, either a generic market belonging to ancillary services market bucket or more often a branch of RCM while

less refer exactly to the combined participation in DAM and IM. In addition, the European market reality offers to the HPS owners the possibility of a wider bidding perspective, giving access to a more lucrative operation strategy through the participation in the IM. The latter perspective is neither studied nor developed in the literature up to this date. More specifically, HPS owners of any market and especially those of the Swiss market should develop their operation strategies on how they would offer their capacity in the Swiss and surrounding market in order to maximise their revenues and in extension to increase the value of their plant in the longterm horizon. Moreover, for the valuation of a plant through the operation simulation, a stochastic problem must be formulated that reveals uncertainty of both DAM and IM prices in a realistic manner while considering more than one scenario in order to avoid over estimation of revenues. 2. Materials and methods

11

prices in the future is designed and implemented based on the three simulated time series (DAM prices proxy, wind generation, solar generation). Furthermore, a hybrid stochastic dynamic programming (SDP) and stochastic linear programming (SLP) optimisation module is designed and implemented in MATLABÒ for simulating the operation of an HPS plant participating in the DAM and IM. The yearly expected revenues is then calculated throughout the horizon of the study. Having calculated the yearly expected revenues’ stream of the plant, specific economic and financial indicators such as the Net Present Value (NPV), the payback time, the Internal Rate of Return (IRR) for the project and the private equity and the Debt Service Cost Ratio (DSCR) are calculated through a financial module designed and implemented within this paper. The flowchart shown in Fig. 1 presents an overview of the developed methodology as it was applied to the specific case study of the Swiss HPS plant participating in the German DAM and IM.

2.1. Methodology structure The developed methodology comprises four (4) distinct modules that are used for calculating the long term value of the HPS plant. Three of these modules are developed in the context of this research while one of them is a commercial tool called AURORAxmp that concerns the DAM simulation. AURORAxmp is a dynamic, long-term, bottom-up electricity dispatch model developed by the EPIS R&D company in the USA, capable to analyse, simulate and optimise various competitive wholesale electricity markets. It adopts fundamental modelling concepts in combination with transmission-constrained dispatch logic and is accompanied by complete sets of databases covering the North American and European electricity territories [29e31]. The developed methodology involves the DAM simulation tool being used in order to generate a fundamentally based marginal cost through simulation of the hourly plants’ operation that will be used as a proxy for the DAM prices. For this purpose, appropriate generation capacity expansion scenarios need to be formulated that represent the respective evolution, under the examined time period, of the power system where the assessed power plant operates. For the scope of the specific case study examined in this paper, such scenarios have been developed based on official EU reports, as it will be explicitly described in the following sections. Through a statistical analysis of the recent years historical prices of the DAM and IM, and the historical values of wind and solar generation, a Multiple Linear Regression (MLR) model for the simulation of a proxy of IM

2.2. Day-ahead proxy market prices simulation According to the literature, in a competitive market, prices should always reflect the marginal cost of production and therefore, be equal with the variable cost of the last generating plant (according to the merit order applied) that is dispatched in order to supply the respective system load demand. The currently applied methodology estimates these hourly market-clearing prices by using an approach that reflects the economics as well as the physical aspects and operational constraints of the power system. More specifically, prices are estimated by applying an appropriate chronological dispatch algorithm that takes into consideration hourly demands, individual resource-operating characteristics, such as their minimum run time and plant start-up costs, as well as the transmission system constraints. The actual dispatch of the generating plant is performed according to their variable cost until the system hourly load demand is supplied. The market-clearing price is calculated according to the cost of meeting an incremental increase in demand and is used for the remuneration of all the generating plant of the system, that are dispatched during the examined time step, for their respective power generation. The valuation of a generation plant requires a number of input parameters that define the operation policy and hence yield the value of this plant. In the context of the specific case study assessed in this paper, the operation of a HPS plant was performed within a liberalised competitive electricity

12

N.A. Iliadis, E. Gnansounou / Energy Strategy Reviews 9 (2016) 8e17

Creation of capacity expansion scenarios for the German power system to 2050 based on the official EU scenarios outlook

Generation of long-term DAM price proxy scenarios for the German market using the DAM simulation tool AURORAxmp

Generation of long-term IM price proxy scenarios for the German market using the developed multivariate information based tool

Long-term operation simulation of an HPS plant participating in the German DAM and IM for the calculation of the expected yearly revenues

Long-term financial simulation of the HPS asset and valuation through economic and financial metrics Fig. 1. Flowchart of the developed methodology.

market, the one in Germany, where the generation plants bid in the DAM and get dispatched accordingly. It was assumed that there was no market power hence the participant was considered as a price taker. The DAM prices proxy was simulated taking into consideration the marginal costs of operation of the plants, the interconnections within the directly interconnected systems to Germany, the technical specifications of the plants, the fuel prices, the demand evolutions of the systems and other relevant technical and economic parameters. For this purpose, the AURORAxmp tool will be used. 2.2.1. Assessed scenarios Two official EU reports [32, 33] were analysed as a source of information and acceptable forecasting for the development of four generation expansion capacity and demand evolution scenarios that were used for the DAM proxy simulation. These scenarios were formed by assuming alternative values on some of the most important features of the electric power systems in the EU Member States as a whole and in Germany in particular. The relative assumptions made concerned mainly the renewable energy sources penetration level, the nuclear power capacity and the implementation of energy efficiency

measures, as they were projected to be established in each year until 2050. An initial Reference scenario (ScR) was assessed while the following three additional scenarios were formed by taking into account assumptions (in comparison to the ones considered in the Reference Scenario) with respect to each country:
High RES scenario e Increased renewable energy sources penetration level (ScH)
Energy efficiency scenario e Additional energy efficiency measures (ScE)
High storage penetration scenario e Increased storage systems penetration (ScRS) It must be noted that the already determined decommissioning plan of nuclear power plants in Germany, which has set very strict targets of complete nuclear phase-out by the end of year 2022, was considered to be the same in all four scenarios assessed. The diagram shown on Fig. 2 presents the annual variations of the German system marginal price obtained for the four scenarios assessed (Reference, High RES, Energy efficiency, High storage) by using the AURORAxmp tool.

It is seen that the High RES scenario having an important RES penetration shows the lowest price in the end of 2050 whereas the three other scenarios show a quite similar evolution. Slightly lower to the Energy efficiency scenario, is observed the High storage scenario where an important capacity of storage was installed. The operation of the storage plant was following the demand in order to cover the peaks using the discharge capacity and the stored energy, and recharge the energy during the low demand hours. This operation has driven the prices to increase slightly with respect to the Reference scenario. This can be explained by the increased demand during the low load demand hours that has driven up the off-peak prices while not decreasing the peak prices. In addition, the energy losses that occur between the discharging and recharging cycle have an impact on the prices. 2.3. Intra-day market prices proxy simulation The IM is a continuously evolving market because of its lack of maturity and because of the changing generation plants mix. The latter shifts the objective of the participants in the way they structure their bidding strategy in this market, where often can be observed patterns of arbitrage between DAM and IM. Moreover, changes occur often enough in the operation process of the RES plants in the market. Starting initially from a plain FIT without any further obligation, going to a dynamic approach where RES operators are responsible for the operation and power balancing of their plants, known also as direct marketing approach. It must be clearly stated that the IM prices proxy simulation is not about the long term forecasting of the IM prices within the time period under study. The purpose is to estimate a parametric model that can be used in order to define the interaction between the IM prices, the DAM prices, the wind generation and the solar generation on an hourly step resolution. For this purpose, appropriate statistical models are usually used and relevant historical data concerning the energy market being assessed are taken into account. Especially the German electricity IM is a new market, which had not started until 2006, presenting a very limited liquidity and interest back then. Although it is being characterised by a constantly increasing interest and liquidity, the time period of these six years can be considered as a short period to reach maturity, in order to provide a homogenous sample appropriate to be used within the context of this paper. Consequently, due to this market immaturity as well as the reasons mentioned above, a long historical time

N.A. Iliadis, E. Gnansounou / Energy Strategy Reviews 9 (2016) 8e17

13

Fig. 2. Marginal price results for the 4 scenarios assessed.

series cannot be taken into account for the specific case study; instead, the historical sample used comprised only the two latest years, 2012 and 2013. However, due to the fact that the expansion capacity scenarios examined for the DAM defined a significantly different generation plants mix along the 35year outlook and among them, the IM prices proxy was not approached through a statistical model. In order to overcome these restrictions, a parametric multivariate model was developed using the information provided by the years 2012 and 2013 and applied it in this 35-year horizon. The dataset’s frequency over this twoyear period is hourly while the following system aspects were considered:
Historical DAM electricity prices [EUR/ MWh]
Historical IM electricity prices [EUR/ MWh]
Historical wind-based electricity generation [MWh]
Historical solar based electricity generation [MWh] The data were provided by EPEX and the Electricity Transmission Operators Amprion, 50 Hz, Tennet and EnBw and consist of 80 784 data points for each sequence in 2012 and 80 760 for 2013, a total of 170 544 hourly values for each variable for the whole 2012e2013 time span. When analysing the respective values for the DAM prices it is seen that a

decrease exists between 2012 and 2013. This decrease can be explained as a combination of the decrease of the system load demand along with the increase of RES in the generation mix. In the same pattern, follow the IM prices, where there was a decrease of the average between 2012 and 2013. The diagram shown on Fig. 3 describes the upper and lower bounds calculated from the developed model. The lower bound represents the 2.5% lower percentile and the upper bound the 97.5% percentile, resulting from the calculation of the prediction interval. 2.4. Hydroelectric pump-storage operation optimisation The problem for the optimisation of the HPS operation is considered multi-stage stochastic regarding the DAM and the IM operation since it is solved 1 day at a time for the DAM and 1 h at a time for the IM. This problem can be approached estimating the expected future revenue, through the Revenue-To-Go function (RTG), and calculate the opportunity costs out of it, the water values. The formulated model can be classified as a stochastic inter-stage problem with stochastic intra-stage sub-problems embedded [20, 34, 35] in a multi-stage and multi-horizon stochastic programming approach [36]. The optimisation is applied for daily time stages where for each time stage water values are calculated considering both DAM and IM. The multi-stage stochastic program was

decomposed into inter-stage and intra-stage sub-problems, with yearly and daily time horizons respectively, and the long term problem is solved in yearly sequences. Especially in Switzerland mixed HPS plants typically have storage reservoirs which are operated seasonally, connected to smaller weekly or daily-operated reservoirs, used for pumping purposes. The future revenue is therefore influenced by short-term decisions either operationally because of e.g. empty daily-operated reservoirs or because of the hourly energy market. At this point it must be noted that the focus of this paper is on capturing the value that arises from the opportunities and arbitrage between the DAM and IM while the disclosure process of DAM and IM prices information is distinct within the model as it is in the real markets too. The decision that needs to be taken concerns the daily water discharge (generation) from the seasonal reservoir 1 and the daily reservoir 2, the pumping from reservoir 2 to reservoir 1, the energy volume to buy and sell in the DAM, and the energy volume to buy and sell in the intraday market, in an hourly resolution. The term hourly resolution is used, because the energy product the algorithm uses in the simulation is for a specific hour in the day, both for the DAM and the IM. Hence, each hour is treated separately, where an energy volume has to be either generated, pumped, bought or sold in the DAM and IM. However, the decision time of the DAM takes place before

14

N.A. Iliadis, E. Gnansounou / Energy Strategy Reviews 9 (2016) 8e17

Fig. 3. Intraday market scenarios upper and lower bound.

the decision time of the IM. In addition, the decision for the DAM is made for all the hours of the day in one step (as it is in the real DAM auction), where the decision for the IM takes place for each hour sequentially (as it is in the real IM trading), knowing the realised prices and decisions before. Furthermore, when the decision for the DAM is made, there is no knowledge neither about the scenario that will be realised in the DAM nor in the IM. In contrast, when the decision takes place for the IM, the DAM scenario realised is known as also all the hourly prices for all the hours in the IM before the hour concerned. 3. Results 3.1. Description The developed methodology was used for assessing the value of a typical HPS plant participating both in the DAM and IM in Germany. As previously mentioned, the selection of Germany has been made on the grounds of the increased market liquidity being obtained compared to other European countries, and on market maturity. For example, in year 2013, France, despite being the second largest energy market in Europe after Germany, had much lower liquidity and volume traded than Germany (the respective values were 11.7% of liquidity and 58 TWh in France and 46.3% liquidity and 246 TWh in Germany), where Switzerland was in the middle (29.3% liquidity and 19 TWh) (European Power Exchange e

EPEX, 2014). It must be noticed that market liquidity is considered to be essential for the proper operation of wholesale markets and the formation of competitive prices that ensure welfare benefits for the end-users. The assessed HPS plant is represented by the simplified diagram shown in Fig. 4 and comprises two reservoirs, four generating units and two pumping units while their detailed technical characteristics are presented on Table 1. The reservoirs are characterised by different cycles, a seasonal for upper reservoir 1 and a daily one for lower reservoir 2. A valid assumption for the hydropower plants in the Alps where the plant is located is that the yearly amount of water inflows remains stable from year to year. Hence, a time horizon of one year was chosen, starting in May, when

the reservoirs are in their lowest level and ending in the end of April the next year. In line with this argument and because the lower reservoir is significantly smaller (1000 times smaller) having a daily operation cycle, even though its contribution to production and pumping is accounted for in the RTG function, its calculation as an additional dimension in the RTG was not accounted for as its impact is negligible and does not affect the global RTG function. 3.2. Operational results The data input in the module were the technical and economic performances of the HPS plant, and eight series of 315,576 simulated values each representing four DAM prices proxy scenarios and eight IM scenarios,

Fig. 4. Simplified diagram of a typical HPS power plant.

N.A. Iliadis, E. Gnansounou / Energy Strategy Reviews 9 (2016) 8e17

Table 1 Technical characteristics of the assessed HPS power plant. Reservoir No.

1 (Upper) 2 (Lower)

Capacity [Mm3]

100,000

100

Turbine No. (twice the same machines)

1, 2

3, 4(*)

Type Max Power Output [MW] Min Power Output [MW] Efficiency (%) Energy Conversion [MWh/1000 000 m3]

Francis 35 15 90 1.1

Pelton 90 8 90 2.7

Pump No. (twice the same machines)

1, 2

Max Power Consumption [MW] 23 Min Power Consumption [MW] 23 Efficiency (%) 90 (*) Turbines 3 and 4 are qualified for delivering of secondary frequency control reserves

totalling 2,524,608 values. The problem consists of 480 two-stage problems per day, summing up to 6,311,520 problems for the days of thirty-five years. The SLP problem describing the IM participation was solved through its deterministic equivalent. For every stage (day) the hybrid SLP/LP twostage problem was solved for every discretised reservoir level (twelve discrete values), for every discretised water discharge (ten discrete values), and for four DAM prices proxy scenarios and eight IM prices proxy scenarios (two per DAM prices scenario). At each stage a RTG function is calculated consisting of twelve points (the same as the reservoir discreet levels), which is used to calculate the previous stage RTG functions, as the solutions advances backwards. The problem was solved on yearly loops, defined according the hydrological cycle of the HPS plant. The sought result of each loop is the RTG function of the first stage for each year, whereas the expected value of the revenues for the entire year is given for different reservoir levels. Since the run starts at the first (1st) of May, the reservoir level is at the lowest, according to the historical data, and therefore the discrete value of level is considered equal to zero (0). The results obtained from two runs assuming that the HPS plant either participates in both DAM and IM or only in the DAM, are presented in Fig. 5. Both of the curves follow the evolution of the four DAM prices proxy scenarios, and could be perceived as their average rate of evolution, since the results are the expected value of revenues considering all four scenarios jointly assuming an equiprobability principle. The total expected yearly revenue in real values, when participating in the DAM and IM, started at 40.1 MEUR and ended in 111.8 MEUR, and when participating in the

DAM only, started at 29.4 MEUR and ended at 79.9 MEUR. As expected the ratio of revenues increase does not change significantly between the expected value in the beginning of the horizon and the end of the horizon, being 36% less for the first and 39% for the latter, showing that the relative level of DAM and IM prices play a secondary role in the operation pattern of an HPS and in the way its flexibility and storage capacity is valued. 3.3. Economic and financial results The results obtained in section 3.2 of the paper were used as input data in order to implement the financial calculation of the HPS plant. For this purpose, a module was developed that takes into account the features of the HPS plant under study in order to assess the financial viability of the project as well as the respective annual revenues, earnings and cash-flows for the time period under study. Furthermore, data concerning the financial aspects assumed for the implementation of the investment project, as well as additional economic data relative to the operation of the plant, were taken into consideration within the developed module. The data considered are presented in Table 2. Based on the assumptions above and considering the expected yearly revenues calculated in the previous section from the operation simulation of the HPS plant in the DAM and IM, the most important financial indexes such as the Net Present Value (NPV), payback time, Internal Rate of Return (IRR), Debt Service Cost Ratio (DSCR) and Levelised Cost of Electricity (LCOE) were calculated. These indexes are commonly applied in the valuation of a power plant but according to the investors’ perspectives and experiences, there might be additional indicators that would suit more in their decision making process. The results calculated for the HPS plant participating both in the DAM and IM and in the DAM only are resumed in the Table 3, where all values are after corporate taxes. 3.3.1. DAM þ IM operation Regarding the operation of the HPS plant in both energy markets, it can be observed, based on the calculation of the NPV, that the project for an initial investment of 217.8 MEUR gives a multiple of 3 in 35 years. The equity payback takes place in 20 years, which is rather common for plants of this value and size, and the project itself has a payback period of 31 years. Even though the Internal Rate of Return can be considered as a misleading and a nonscientific financial indicator, for the private equity is 12.8%, a number considered to be high for such a size of an investment, and for the project 7.9%, where the two are related through the share between the private equity

15

and the debt. The higher the leverage the higher is the equity IRR and the lower is the project IRR. For 100% of private equity the project IRR is equal to the private equity IRR. 4. Conclusions The purpose of this paper is to present the main aspects of a methodology that has been developed in order to be used for the long term valuation of a HPS plant, participating in the DAM and IM. Within this scope and framework, a number of different subjects were researched through the analysis of a case study that concerns the valuation of a typical Swiss HPS power plant active in the DAM and IM German market. The main issues being effectively addressed concern the following issues:
How the price proxy scenarios for the DAM be calculated for a horizon of thirtyfive (35) years?
How the IM prices proxy scenarios can be simulated for a horizon of thirty-five (35) years?
Which is the optimal simulation of allocation and combination of power output under uncertainty destined to the DAM and IM respectively, in order for the utility to achieve maximum expected revenues and thus be used to capture its long term value? For the consistent and realistic valuation of a plant in the long term, which sources a large part of its revenues from the activity in the short term markets of DAM and IM, an appropriate model was designed, formulated and implemented. The results were the expected yearly revenues of the HPS plant by maximising the revenues from generating, pumping, buying and selling in the DAM and IM. The stochastic approach delivers realistic results considering in the solution, simultaneously various DAM and IM prices scenarios that might occur in the future. The valuation of the HPS plant, taking into account a thirtyfive year (35) years horizon, requires scenarios that have a technical and economic significance as the ones used within this paper. The yearly expected values being calculated were fed into an economic and financial analysis module, developed within the framework of this research, for the calculation of a series of indicators such as the NPV, the payback time, the IRR for the project and the private equity and the DSCR. A number of assumptions were taken into consideration within the model such as the Capital Expenses, the financing structure of the project, the cost of equity, the duration of the investment and the lifecycle of the project, the interest rate and tenure, the corporate

16

N.A. Iliadis, E. Gnansounou / Energy Strategy Reviews 9 (2016) 8e17

Fig. 5. Comparison of revenues of HPS for participating in the DAM and IM vs. DAM.

tax coefficient, the Operation Expenses, and the accounting amortisation scheme. The results calculated were the expected revenues per year in different balance sheet stages, their graphs and the indicators mentioned above. The results for the HPS plant participating in the DAM and IM showed after corporate tax, an NPV of 653 MEUR, a payback of the in thirty-one (31) years and of private equity in twenty (20) years, an IRR of the project of

Table 2 Financial and economic assumptions for the financial calculation of the HPS plant. Field

Units

Value

Installed capacity Project life duration Overnight cost Private equity Debt Taxes coefficient Bank reserve coefficient Yearly amortisation Private equity cost Debt interest rate Discount rate Investment duration Loan type Tenure Goodwill period Inflation Fixed Operation and Maintenance Variable Operation and Maintenance

[MW] [years] [EUR/MW] [%] [%] [%] [%] [%] [%] [%] [%] [years] [e] [years] [years] [%] [EUR/MW/y]

245 80 4,444,880 20 80 14 5 3 5.5 3 3.5 35 Mortgage 35 0 1.5 13,850

[EUR/MWh]

4.14

7.9% and of the private equity 12.81% (for 20% of private equity) and a minimum DSCR of 0.84. Nevertheless, even though almost 40% of the revenues came from the flexibility of the plant to extract the value by participating in the DAM and IM, the level of the prices will play an important role to the remaining 60% of the value. Comparing to the results above, a HPS participating in the DAM only, showed deteriorated numbers after corporate tax, having an NPV of 243 MEUR, a payback of the project in thirty-four (34) years and of the private equity in twenty-eight (28) years, an IRR of the project of 6.5% and of the private equity of 10.32 (for 20% of private equity) and a minimum DSCR of 0.59 and an LCOE of 27.6 EUR/MWh. The NPV had a significant drop even though the result is diluted because of the long horizon of eighty (80) years. The IRR, both for the project and the private equity, showed a small difference in percentage

points, and one might ask, looking only at this result, why should the company enter into the extra risk of participating in two markets. There was an important impact on the project in terms of DSCR, rendering the project even more difficult for an institution to finance it without further private equity injection and corporate guarantees. To sum up, it can be safely deducted that an HPS plant has important flexibility because of its fast response times and storage capacity to participate in both the DAM and IM and cease the opportunities between them. Considerations in the design of such a plant have to be made in terms of the lower reservoir design, size of pumps, high endurance in multiple start-ups per day, low responsiveness to achieve maximum capacity and rapid switching between regimes of operation, in order to maximise its flexibility. Additional care should be taken when defining the risk limits concerning this plant, as a very risk

Table 3 Financial and economical results from the financial calculation of the HPS plant when participating in both DAM and IM and DAM only. Field

DAM þ IM operation

DAM operation

Net Present Value (NPV) [MEUR] Payback of private equity [years] Payback of project [years] Internal Rate of Return (IRR) of the private equity [%] Internal Rate of Return (IRR) of the project [%] Minimum Debt Service Cost Ratio (DSCR)

2867.9 20 31 12.8 7.9 0.84

1809.9 28 34 10.3 6.5 0.59

N.A. Iliadis, E. Gnansounou / Energy Strategy Reviews 9 (2016) 8e17

prone approach might expose the company in unacceptable risk levels and a very risk averse approach might limit the revenue stream of the plant.

[5]

[6]

Acknowledgments We would like to thank Loukas DAOUTIS, Hubert ABGOTTSPON, Stein-Erik FLETEN, the EPIS team, Deborah AUSTIN-SMITH and Ryan SWARTZ. Abbreviations

CapEx DAM DSCR EPC EPEX EST HPS IM IRR MLR NPV OpEx RES RTG SDP SLP

Capital Expenses Day Ahead Market in electricity Debt Service Cost Ratio Engineering Procurement Construction European Power Exchange Energy Storage Technologies Hydro Pumped-Storage Intraday Market in electricity Internal Rate of Return Multiple Linear Regression Net Present Value Operational Expenses Renewable Energy Sources Revenue to Go Function Stochastic Dynamic Programming Stochastic Linear Programming

[7]

[8]

[9]

[10]

[11]

[12]

[13]

[14]

[15]

[16]

References [1] S.-F. Zhan, X.-C. Zhang, C. Ma, Coal classification based on environmental protection and burning quality, J. China Coal Soc. (2009). [2] X. Zhang, L. Cao, K. Caldeira, Energy switching threshold for climatic benefits, in: AGU Fall Meeting, 2013. [3] X. Zhang, W. Chen, C. Ma, et al., Assessment method for regional environmental risk based on pressure-state-response model, China Environ. Sci. 32 (2012) 84e87. [4] X. Zhang, N. Myhrvold, K. Caldeira, Key factors for assessing climate benefits of natural gas versus

[17]

[18]

[19]

coal electricity generation, Environ. Res. Lett. (2014). S. Zhan, X. Zhang, W. Chen, Dynamic modelling for ecological and economic sustainability in a rapid urbanizing region, Procedia Environ. Sci. 13 (2012) 242e251. C. Ma, M. Ju, X. Zhang, H. Li, Energy consumption and carbon emissions in a coastal city in China, Procedia Environ Sci. 4 (2011) 1e9. X. Zhang, W. Chen, C. Ma, S. Zhan, Modeling the effect of humidity on the threshold friction velocity of coal particles, Atmos. Environ. 56 (2011) 154e160. X. Zhang, W. Chen, C. Ma, S. Zhan, Modeling particulate matter emissions during mineral loading process under weak wind simulation, Sci. Total Environ. 449 (2013) 168e173. ´ . Baı´llo, S. Cerisola, J.M. Ferna A ´ndez-Lo ´pez, R. Bellido, Strategic Bidding in Electricity Spot Markets under Uncertainty: a Roadmap, 2006. A.J. Conejo, M. Carrion, J.M. Morales, Decision Making under Uncertainty in Electricity Markets, in: F.S. Hillier (Ed.), Springer, 2010. M. Thompson, M. Davison, H. Rasmussen, Valuation and optimal operation of electric power plants in competitive markets, Oper. Res. (2004) 546e652. M. Pedram, S. Moff, A stochastic programming framework for the valuation of electricity storage, Cleveland (2012). A. Botterud, Evaluation of investments in new power generation using dynamic and stochastic analyses, in: 8th International Conference on Probabilistic Methods Applied to Power Systems, Iowa, 2004. C. Weber, O. Woll, Valuation of CHP power plant portfolios using recursive stochastic optimization, in: International Conference on Probabilistic Methods Applied to Power Systems, 2006. PMAPS 2006, Stockholm, 2006. S.N. Siddiqi, Project valuation and power portfolio management in a competitive market, IEEE Trans. Power Syst. 15 (2000) 116e121. S.-J. Deng, Stochastic Models of Energy Commodity Prices and Their Applications: Mean-reversion with Jumps and Spikes, POWER working paper, University of California Energy Institute, Berkeley, 1999. R. Deb, Operating hydroelectric and pumped storage units in a competitive environment, Electr. J. 13 (2000) 24e32. J. Pinto, J. de Sousa, M.V. Neves, The value of a pumping-hydro generator in a system with increasing integration of wind power, in: 8th International Conference on the European Energy Market (EEM), Zagreb, 2011. P. Kanakasabapathy, K. Shanti Swarup, Bidding strategy for pumped-storage plant in pool-based electricity market, Energy Convers. Manag. 51 (2010) 572e579.

17

[20] N. Loehndorf, D. Wozabal, S. Minner, Optimizing the Day-ahead Bidding Strategy of Electricity Storage Using Approximate Dynamic Programming, 2010. [21] D. Wilde, How Can Pumped-storage Hydroelectric Generators Optimise Plant Operation in Liberalised Electricity Markets with Growing Wind Power Integration?, 2010. [22] N. Lu, J.H. Chow, A.A. Desrochers, Pumped-storage hydro-turbine bidding strategies, IEEE Trans. Power Syst. a Compet. Electr. Mark. vol. 19 (2) (May 2004) 834e841. [23] N. Loehndorf, S. Minner, Optimal day-ahead trading and storage of renewable energiesdan approximate dynamic programming approach, Energy Syst. 1 (1) (2010) 61e77. [24] W.B. Powell, Approximate Dynamic Programming, Wiley, Princeton, 2011. [25] M. Olsson, L. So ¨der, Hydropower planning including trade-off between energy and reserve markets, in: 2003 IEEE Bologna Power Tech Conference, Bologna, 2003. ´ . Plazas, A.J. Conejo, F.J. Prieto, Multimarket [26] M.A optimal bidding for a power producer, IEEE Trans. Power Syst. 20 (2005) 2041e2050. [27] C. Corchero, F. Herredia, E. Miijangos, Efficient solution of optimal multimarket electricity bid models, in: 8th International Conference on the European, Energy Market (EEM), 2011, pp. 244e249. [28] R. Schuman, Market Design with Wind: Managing Low-predictability in Intraday Markets, European University Institute, Florence, 2012. [29] S. Carley, Decarbonization of the U.S. electricity sector: are state energy policy portfolios the solution?, 2010. ´ . Gallacho [30] A.M. Foley, B.P.O ´ir, J. Hur, R. Baldick, E.J. McKeogh, A strategic review of electricity systems models, Energy 35 (12) (2010) 4522e4530. [31] AURORAxmp, 2012 [Online]. Available: http:// www.epis.com [Accessed 2012]. [32] European Commission, Energy Roadmap 2050 e Impact Assessment and Scenario Analysis, 2011. [33] European Union, EU Energy, Transport and GHG Emissions e Trends 2050, Reference Scenario 2013, December 2013. [34] G. Pritchard, A. Philpott, P. Neame, Hydroelectric reservoir optimization in a pool market, Math. Program. 103 (2005) 445e461. [35] H. Abgottspon, G. Andersson, Approach of integrating ancillary services into a medium-term hydro optimization, in: XII SEPOPE: Symposium of Specialists in Electic Operational and Expansion Planning, 2012. [36] M. Kaut, A. Werner, K. Midthun, A. Tomasgard, L. Hellemo, M. Fodstad, Multi-horizon stochastic programming, Computational (2013) 1e15.