Renewable Energy 77 (2015) 227e239
Contents lists available at ScienceDirect
Renewable Energy journal homepage: www.elsevier.com/locate/renene
Economic risk analysis of decentralized renewable energy infrastructures e A Monte Carlo Simulation approach € Uwe Arnold a, 1, Ozgür Yildiz b, * a b
€t Weimar, Faculty of Civil Engineering, Department of Urban Infrastructure Management, Germany AHP GmbH & Co. KG/Bauhaus-Universita €t Berlin, School of Economics and Management, Department of Economic Policy and Environmental Economics, Germany Technische Universita
a r t i c l e i n f o
a b s t r a c t
Article history: Received 23 October 2013 Accepted 18 November 2014 Available online
There are several different economic barriers such as high up-front capital costs, high transaction costs and divers risks (e.g. performance and technical, contract risks, market risks) that keep potential investors or institutional lenders from investing in decentralized renewable energy technologies (RETs). Therefore, suitable business models, specific financing concepts and advanced risk management tools to deal with issues concerning transaction costs and financial risks are required to support RET investments. This article deals with this issue by introducing a Monte Carlo Simulation (MCS) approach to risk analysis based on an entire life-cycle representation of RET-investment projects. By doing this, the authors uncover considerable advantages regarding content and methodology compared to ordinary NPVestimation or sensitivity analysis. It could be shown that the presented financial analysis combined with MCS aids in optimizing the conceptual design of an investment project with respect to capital returns and risk. Since both issues are decisive for lenders and investors, the double-criteria analysis method presented in this paper facilitates the raising of capital for project investments in decentralized RETs. © 2014 Elsevier Ltd. All rights reserved.
Keywords: Renewable energy technologies Monte Carlo Simulation Risk analysis Simulation applications Financial engineering Bioenergy
1. Introduction A shift towards renewable energy sources characterizes economic and environmental policy measures in countries all over the world. Especially in the European Union (EU), ambitious goals for climate change mitigation have been set. According to this, mandatory national overall targets and measures for the use of energy from renewable sources are defined by the European Commission (Directive 2009/28/EC) to increase the share of renewable energy in the EU gross final consumption of energy to a minimum of 20% by 2020 [1]. Among other strategies pursued in most countries to reach this ambitious goal, decentralized, small scale renewable energy technologies (RETs) and the related energy production facilities, such as small scale bioenergy infrastructures, small scale wind farms and
€t Berlin, Fakulta €t VII e Wirtschaft * Corresponding author. Technische Universita €konomie und Wirtschaftspolitik, (Sekreund Management, Fachgebiet Umwelto tariat H 50), Strabe des 17, Juni 135, 10623 Berlin, Germany. Tel.: þ49 30 31425264. E-mail addresses:
[email protected] (U. Arnold),
[email protected] € Yildiz). (O. 1 AHP GmbH & Co. KG, Karl-Heinrich-Ulrichs-Straße 11, 10787 Berlin, Germany. web: www.ahpkg.de. http://dx.doi.org/10.1016/j.renene.2014.11.059 0960-1481/© 2014 Elsevier Ltd. All rights reserved.
solar plants, are changing the structure and characteristics of the regional, national and even interconnected international energy supply infrastructures to an ever more rapidly growing extent. In spite of numerous benefits resulting from the implementation of RETs, e.g. their contribution to environmental protection, their impacts on economic growth by creating jobs, by forming human capital and by offering a market for new business models, several technical and economic barriers delay the implementation of RETs. Among these, financial obstacles play a particular role [2e5]. First of all, RETs usually come with higher power-specific upfront capital costs than investments into conventional energy infrastructures. Furthermore, high transaction costs and other risks (e.g. performance and technical risks of the used technical facilities, contract risks with the suppliers of raw materials such as bioenergy crops, market risks such as future price developments and the impact of demographic changes to the local demand of energy) may hinder potential investors or institutional lenders to invest in RETs [6,7]. In general, capital costs are a function of the borrower's credit rating, the provided securities, the leverage ratio, and the aggregated project risk. Usually, higher aggregated project risk leads to higher interest rates requested for the loan or even to the complete loan denial by lenders such as banks. The spread of interest rates for
228
€ Yildiz / Renewable Energy 77 (2015) 227e239 U. Arnold, O.
higher risk investments is one of the direct consequences of the Second and Third Basel Accords of the international regulatory system for banking. Equity investors and institutional lenders link their return on investment (ROI)-expectations with project inherent risk. Thus, as in other domains of asset-financing, risk assessment and risk management are crucial prerequisites for financial feasibility of renewable energy projects. Therefore, coping with financial constraints of renewable energy technology (RET)-investments requires a stable and reliable political and legal framework so potential investors can reduce regulatory risks and hence significantly reduce the cost of capital [8]. Moreover, suitable business models, specific financing concepts and advanced risk management to deal with transaction cost and financial risk issues [9] are required to support RET investments under undistorted market conditions (i.e. with gradually reduced subsidies and shrinking acceptance of public risk coverage for economic, technological or political reasons). This article addresses the necessity of advanced risk management tools, and presents a Monte Carlo Simulation (MCS) approach to risk analysis based on a representation of the entire life-cycle of RET-investment projects. Therefore, this paper is structured as follows: Section two briefly describes characteristics of risk and risk management within the framework of RETs. Subsequently, the structure of the developed models and analysis tools, including MCS as a tool to evaluate the risk of an investment in RETs, are introduced and specified for the case of a wood heating plant. This basic introduction to the methodology of risk oriented financial analysis of investment projects is then demonstrated with an €fenhainichen, application case example of a planned project in Gra Germany in section four. The paper closes with some concluding remarks and an outlook on future developments. 2. Risk analysis in decentralized renewable energy projects Risk management of decentralized renewable energy projects consists of a sequence of different measures to identify, assess and allocate project risks. The aim of this procedural chain is to focus attention at potential factors that could have an impact on project cash flows, to analyze qualitatively and quantitatively the possible effects of an adverse event on project earnings and consequently on its viability, and finally to reduce risks by adopting appropriate measures within the project company or by delegating a specific risk to a third party. A minimum requirement an appropriate risk management system must fulfill from a lenders' point of view is the principle that the project should be able to cover debt service with its cash flows even in a worst-case scenario. Therefore, lenders resort to key figures such as the so called debt service cover ratio (DSCR), which determines a projects capability to cover debt servicing from its cash flows, to evaluate a project. From an investor's perspective, the objective of risk management is to assure that the project is able to generate a proper return on equity in a base-case scenario which corresponds to the incorporated risk. For this purpose, fundamental key figures such as the internal rate of return (IRR) or the net present value (NPV), which convey a project's attractiveness to investors, are subject to risk analysis [10]. The individual risks which have an impact on a project's cash flow can be divided into different categories: Pre-completion phase risks, post-completion phase risks and issues common to both phases. Within the scope of pre-completion risks are primarily technical and construction risks whereas supply, operational and market risks constitute mainly post-completion risks. Risks arising from financial, legal, regulatory or environmental spheres build the group of risks common to both phases. Coping with these individual risks is achieved via various measures such as so called turnkey contracts for construction risks, take-or-pay and bring-or-
pay agreements or other contractual agreements for market risks and e.g. insurance policies for environmental and operational risks, to name but a few [11]. Nevertheless, risk positions stay relevant even with these measures so that there is a need for investors and lenders to assess the remaining entrepreneurial risk. Therefore, in the following a MCS approach is presented. 3. A Monte Carlo Simulation approach to risk analysis e the method and toolkit So far, Monte Carlo techniques have rarely been used within the context of risk management of renewable energy infrastructures as they require considerable data processing and the definition of probability density functions for random input variables such as fuzzy or uncertain design and forecast parameters. Nevertheless, some examples exist such as MCS applications to wind energy [12,13] or within the context bioethanol production [14]. The approach presented in this paper focuses on bio-energy infrastructures and analyzes the financial risk for investors and lenders by subjecting the NPV of bio-energy projects to an MCS. Prior to describing the methods chosen, some definitions are necessary. The project-NPV yNPV,0 is defined by equation (1)
yNPV;0 ðT; iD Þ ¼
T X
CFtot;t ð1 þ iD Þt
(1)
t¼1
with yNPV,0: net present value related to t ¼ 0 CFtot,t: total cash flow (CF) in period t including investment, operational, financing cash flows and the liquidation revenue for t¼T iD: discount interest rate (measure of opportunity cost) T: number of balance periods (e.g. years) A zero-NPV indicates that investors receive complete repayment of the capital invested plus an appropriate interest according to the discount interest rate iD. A negative NPV denotes that the investment cannot generate sufficient returns in order to compensate for opportunity costs. Positive NPV-values classify investment projects to be above average expectations of profitability. The investor expectations with respect to proper interest on equity are represented by the discount interest rate iD which influences the NPV-value significantly and serves as an indicator of opportunity cost. The NPV was chosen as a major comprehensive measure of financial feasibility and project profitability since it is easy to understand, convincing and practical, even for those (potential) project participants with little background in investment analysis [15]. This accessibility is of particular importance within the financial analysis of decentralized renewable energy infrastructures where local stakeholders are often active drivers for the realization of infrastructure projects, but lack any investment analysis knowledge and therefore need practical tools to handle and overcome this constraint [16]. The financial project risk R of negative project NPVs may be defined as:
Z0 RNPV < 0 ðx1 …xn ; iD Þ ¼
~NPV;0 ~NPV;0 dy pdf y
(2)
∞
with RNPV1000) of the previous sequence of steps create collections of random samples for each of the interesting output parameters which can be described by their pdfs and statistical properties (mean value, standard deviation etc.) as well. 3.1.1. Determination of probability density functions of MCS input variables The crucial step of MCS, however, consists in the determination of the required probability density functions for the random input parameters xi. The (input-) pdfs of the random model input variables xi must be already known in order to apply a Monte Carlo Simulation as described. If a data based statistical derivation of the required pdfai (xi) is not possible due to the lack of suitable data, a knowledge based empirical approximation should be used instead. In the application cases presented within this paper, the pdfai (xi)derivation procedure, which is illustrated in Fig. 2, was applied. For each of the MCS-relevant input quantities xi a first analysis is carried out based upon literature, reference cases and expert interviews revealing variation ranges of the input quantities of interest. Afterwards, the input values xi are discussed in a workshop with the stakeholders involved in the project in order to determine plausible worst case, base case, and best case estimates. Usually, practitioners have sufficient experience and significant intuition to specify confidence intervals [worst case value; best case value] for each single input quantity. These empirical values of the random input quantities can be interpreted as quantiles q(p) of a probability density function. Based upon these quantile-values, a specification of the type of the distribution
Fig. 1. Operating principle of a Monte Carlo Simulation (source: author's design, adapting sources [17,18]).
€ Yildiz / Renewable Energy 77 (2015) 227e239 U. Arnold, O.
230
Fig. 2. Derivation of probability density functions pdfai of MCS-input variables xi from anchor values of scenario analysis (source: author's design, adapting sources [17,18]).
(normal, log-normal) and of the corresponding probability-value p of the quantile, the pdfai (xi) can be calculated by means of the following equations:
quantile value q(p) and specifying the related probability p (e.g. 5% or 95%) leads to the standard deviation s. 3.1.1.2. Log-normal distribution. pdf(x):
3.1.1.1. Normal distribution. pdf(x): 2 2 1 f ðxi Þ ¼ pffiffiffiffiffiffiffieðxmÞ =2s s 2p
flog ðxi Þ ¼ (3)
2 1 pffiffiffiffiffiffiffieðln xi mÞ xi s 2p
1 Fðxi Þ ¼ pffiffiffiffiffiffiffi s 2p
Zxi Flog ðxi Þ ¼
ðxmÞ2 =2s2
e
dx
(4)
∞
∞
with Eðxi Þ
qðpÞ m s¼ F1 p
; xi > 0
2 1 pffiffiffiffiffiffiffieðln xmÞ xi s 2p
2s2
dx
(8)
2 mean value; expectation value : Eðxi Þ ¼ emþs =2
m: m ¼ ln[E(xi)]s /2 2
b s : standard deviation; b s ¼ emþs
(5)
2
s from quantile q: for ln x m:
sn
with q(p): quantile value for probability p with xi < q F1 sn (p): inverse function of f(x) for standard normal distribution
rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi pffiffiffiffiffiffiffiffi p ¼ 2 ln p 2p F1 sn
(7)
(9)
with m: mean value; m ¼ E(xi): expectation value s: standard deviation; s2: variance ¼ VAR(xi)
s from quantile qðpÞ:
2s2
cumulative distribution function:
cumulative distribution function:
Zxi
(6)
Using the base case value of xi as the expectation value E(xi), selecting one of the extreme values (worst case or best case) as the
for ln x > m:
ln qðpÞ m s¼ 1 p F1 sn
=2
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi 2 es 1
ln qðpÞ m s¼ F1 sn p
(10)
(11)
Using the base case value of xi for E(xi), selecting one of the extreme values (worst case or best case) as the quantile value q(p) and specifying the related probability p (e.g. percentiles 5% or 95%) leads to m and s by means of the iterative solution of the equations system (9)e(11). For more details see Ref. [19]. The decision whether to select a normal or a log-normal distribution type for the input variable xi is facilitated by the answer to
€ Yildiz / Renewable Energy 77 (2015) 227e239 U. Arnold, O.
the question whether negative values of xi can occur. In many cases where negative values of xi are not possible the log-normal distribution type is more appropriate than that of a normal distribution. 3.1.2. Selection of random MCS-input variables from sensitivity analysis Both, the iterative determination of MCS-input pdfs and the MCS-sequence itself, create significant computational processing effort which is a function of the number of random input variables xi. Therefore, with respect to computation time economics in practical applications, it makes sense to constrain the number of MCS-input variables xi to those with the highest impact on the model-output yj. Prior to carrying out the MCS-analysis, a sensitivity analysis of the deterministic model is used to rank the model input parameters xi according to their relevance for the sensitivity of the model jvyj/vxij. Fig. 3 illustrates a characteristic result of the sensitivity analysis carried out for the deterministic model of the wood-based bio-energy plant described in chapter 5. Within the sensitivity analysis of the deterministic model, a single input parameter is varied systematically within a predefined range of values [xi,min;xi,max]. All the other input variables xksi are kept constant with their base-case value. The related deviations of the model-output variables of interest yj (here: project NPV) from its base-case value are recorded generating a sensitivity characteristic of the model yj =yj ¼ Si ðxi =xi Þ as displayed in Fig. 3. High gradients of the different characteristic curves Si in Fig. 3 indicate high model sensitivity to variations of the related input parameter. In the example of Fig. 3, for instance, the standardized model sensitivity to small relative changes of fuel price, total plant investment, repair & maintenance cost rate and fuel humidity is above average with vðyj =yj Þ=vðxi =xi Þ > 1. 3.1.3. MCS-output: financial risk indicators The application of the MCS provides detailed insight into the probability distribution of the target variable (here: project NPV). In
231
concrete terms, features such as the mean value or the median, left, right- or two-sided confidence intervals, and the cumulative risk of negative values (insufficient return on equity (ROE) or even equity loss) can be calculated (see Fig. 4). This analytically generated information helps potential investors and lenders to grasp the inherent risk of an investment and to expand the possibilities of plant and project design optimization. Fig. 4 displays the probability density function pdf(yNPV,0) of the project-NPV as a histogram and shows the portions of the NPVdistribution below and above the discriminating value of yNPV ¼ 0. The area to the left of the yNPV ¼ 0 line represents the financial risk RNPV
