EFFICIENCY ANALYSIS OF TURKISH POWER PLANTS USING DATA ENVELOPMENT ANALYSIS

Transcription

1 EFFICIENCY ANALYSIS OF TURKISH POWER PLANTS USING DATA ENVELOPMENT ANALYSIS I. Or and K. Sarica Industrial Engineering Department, Bogazici University, Istanbul, Turkey; ABSTRACT Performance of electricity generation plants in Turkey are analyzed and compared. The data set contains inputs from 65 thermal, hydro and wind power plants, owned by the private and public sectors. Data Envelopment Analysis (DEA) is used as the primary mathematical tool. Two efficiency indexes, reflecting operational and investment performance, are defined and pursued. Constant returns to scale, variable returns to scale and assurance region type DEA models are used in the analysis. Scale efficiency is also considered. Performance comparisons include public versus private sector plants, natural gas versus coal versus oil fired plants. Also, relationships between efficiency scores and various input/output factors are investigated and some interesting trends are identified. Key Words: DEA, Decision Modelling and Theory, Energy Systems Planning. 1. INTRODUCTION In electric power systems one of the key decisions is the type of power plants that should be installed. Currently in Turkey renewable and thermal plants are the two main categories. Renewables are sub classified into hydro power plants (HEPP), wind, solar, renewable combustibles and geothermal. Thermals are sub classified as coal, natural gas and oil fired power plants [IEA(2001)]. The question that arises is Which one is better and why?. The answer not only depends on the worldwide supply and demand profile of primary energy resources, but is also highly influenced by global trends and technological developments. Many studies involving comparisons among types of power plants, have been made in order to obtain the best answer to this question. Most of these studies are based on comparisons of power plants with respect to thermal efficiencies, while economic evaluations of plants are made upon monetarization of thermal inputs and outputs like fuel and electricity [Bakos (2003), Kwak et al (2003), Liu et al (2003), Park et a l(2000), Sàez et al (1998)]. This study involves a multi input-multi output performance evaluation and comparison of a set of power plants in Turkey based on real data. A key issue in any such multi factor study is the determination of an objective and realistic set of weights on the inputs and outputs. In this study, this issue is handled within the framework of an appropriate mathematical tool, namely Data Envelopment Analysis (DEA), which was introduced by Charnes et al (1978). Accordingly, two DEA models (one focusing on annual operational performance and the other on long term investment performance efficiency) are defined for the power plants investigated. Evaluations and comparisons within the framework of these models give consideration to various cost, resource availability, production and environmental factors. The quantification (and monetarization) of environmental effects is done based on an earlier project supported by the European Commission, named ExternE Project. The results of the

2 study not only provide a general efficiency ranking and evaluation of the investigated power plants, but also facilitate various interesting efficiency comparisons, such as public versus private sector plants, coal versus natural gas plants, renewable versus thermal plants. Furthermore, relationships between efficiency scores and various input/output factors are investigated and some interesting trends are identified. 2. MODEL DESCRIPTION As indicated, the aim has been to develop mathematical models and approaches which will facilitate efficiency / effectiveness comparisons of electric energy generation facilities. For this purpose, two basic models are formed. One is related to the annual operational performance of the targeted facilities and other is related to their long term investment performance Models to Evaluate Operational Efficiency of Power Plants In this context, two Operational Efficiency Models have been developed: one for thermal power plants, the other for renewable source power plants. The basic DEA model for the operational performance of thermal power plants consists of 6 parameters. These are: fuel cost, production, availability (utilization), thermal efficiency, environmental cost, and Carbonmonoxide (CO). The first three parameters are chosen in order to give due consideration to the primary mission of a power plant (which is reliable generation of electric energy at an acceptable cost). The last three parameters are chosen to reflect the key societal condition imposed on power plants (which are related to their effects on the local and global environment). Fuel cost is the total annual cost of the input fuel used in the plant (in U.S. $). It is chosen as an input since it is one of the main resources used for production and its inclusion basically eliminates the necessity to consider other monetary factors in the model at operational level. Production reflects the annual amount of electricity produced by the power plant (in kwh). It is chosen as an input since it provides a good quantitative measure of the primary output of a power plant. Utilization is the annual percentage of time a power plant is actually producing power. Here the expectation is to have a measure of the plant s availability and reliability for electricity production and delivery. It is taken to be an output since an increase in utilization is an important positive sign. Thermal efficiency is the average annual efficiency that relates how much of the heat dissipated is converted into electric energy; it reflects the effects of CO 2 emissions. It has environmental and economic implications such that maximization of efficiency minimizes emissions and fuel consumption simultaneously. No unit is given since it is a ratio. Environmental cost has been included in order to take into account the negative environmental effects of power plants. It is derived from the annual SO 2, NOx and particulets emission (in tones) of each plant. The negative effects are then converted to monetary units using the $ per ton conversion value obtained from ExternE Project, Finland results [ExternE (1995)] CO is selected since its emissions have serious environmental and health effects, while it is not covered by the ExternE project. It is also a measure of burning quality of plants. Like emissions, it is measured in tones per year. The DEA model developed for the operational performance of renewable power plants has one input factor (operating cost) and two output factors (production and utilization). These three parameters are chosen in order to give due consideration to the primary missions of a power plant (which is reliable generation of electric energy at an acceptable cost). Since

3 environmental effects addressed and measured by the factors deployed in the thermal plants model have negligible values in renewable source plants, environmental effects of renewable source plants are not considered. A new factor included in the renewable power plant operational performance model is operating cost, which is used in place of fuel cost for thermal power plants (since there is no direct fuel cost in this case). It includes annual maintenance and operational costs Models to Evaluate the Long Term Investment Performance of Power Plants The DEA model developed for the evaluation of long term investment performance of power plants has 4 parameters. These are investment cost, installed power capacity, construction time and average utilization. These parameters are chosen to reflect the overall effectiveness of an electricity generating facility, during its full economic life. As such, their values should be representative of the full economic life of the facility. Investment cost is the total funds used for the construction of the power plant from the design phase up to the start of operations. It is treated as an input since a decrease in investment amount (while keeping other factors unchanged) is a positive sign. Construction time is the time between start up of construction and initiation of electricity production. It is treated as an input since a decrease in construction time (while keeping other factors unchanged) is a positive sign. Power is an important design parameter reflecting the potential amount of electricity that can be produced in a unit of time (units deployed are MW). It is treated as an output since an increase in power (while keeping other factors unchanged) is a positive sign. Utilization (percentage) is the average utilization period of a plant over 8 years (eight year period is considered because of the limited data available). It is expected to characterize the long-term availability and reliability of the plant. It is treated as an output since an increase in utilization (while keeping other factors unchanged) is a positive sign. 3. DATA COLLECTION AND COMPILATION Throughout this study data collection has been a crucial, sensitive and time consuming effort. There are two reasons for this: i) the difficulty in obtaining correct data because of strategic and/or secrecy considerations of the source institutions, ii) the sensitivity of the DEA) to data values (a small change/error having the potential to cause a big change in the result obtained). Because of this sensitivity and difficulty, data is gathered through various sources. Data for publicly owned thermal power plants are gathered through the EÜAŞ (Electricity Production Company) and Ministry of Energy. Data for publicly owned hydro power plants are gathered through DSİ (The State Hydraulic Works). Data regarding privately owned power plants is compiled through the aggregation of fragmented pieces from different companies. In this regard, data is obtained from 3 different types of companies: natural gas fueled CC type autoproducers, wind energy fueled BO (Build Operate) producers, natural gas fueled CC type BO producers. Detailed information (such as name or location) is not given regarding the individual power plants investigated. The only ownership information that is provided is whether the plant is privately or publicly owned. Unfortunately, investment costs of the power plants do not include a present value (PV) analysis (missing information such as date, number and amount of payments has rendered PV computations impossible). They are, however, quoted in U.S. $ units (which has a lower discount rate than Turkish Lira), thus reducing deviations from a PV analysis. 3.1 Data Pre-Processing The data obtained from the public and private plants investigated are combined under 4 sets: i) investment performance data set of thermal power plants, ii) operational performance data

4 set of thermal power plants, iii) investment performance data set of renewable power plants iv) operational performance data set of renewable power plants. The first two sets are displayed in Tables 1 and Table 2 respectively. The environment cost values in these data sets are the dollar equivalents of the related plants SO 2, NOx and particulets emissions with respect to emissions/cost relationships obtained from ExternE project for Finland. The associated cost of air pollutants per ton used through out this study are: SO 2 : $ 1,250 /ton; NOx: $ 1,100 /ton; Particulets: $ 2,000 /ton Table 1. Investment Performance Data Set for Thermal Power Plants Plant Investment Cost ($) Construction Time (Month) Power (MW) Working Hours Public 1 1,719,419, ,360 4,528 Public 2 322,349, ,003 Public 3 644,389, ,474 Public 4 304,898, ,3 4,851 Public 5 1,053,960, ,465 Public 6 419,602, ,656 Public 7 98,792, ,122 Public 8 496,840, ,571 Public 9 402,394, ,692 Public ,482, ,830 Public ,137, ,665 Public ,886, ,302 Public ,143, ,350 5,092 Public ,400, ,410 7,635 Public ,754, ,623 Private 1 440,000, ,200 Private 2 780,000, ,540 8,200 Private 3 820,000, ,520 8,200 Private 4 27,297, ,5 8,116 Private 5 40,595, ,225 Private 6 58,315, ,447 Private 7 4,286, ,3 8,087 Private 8 10,297, ,336 Private 9 4,076, ,625 Private 10 9,828, ,118 Private 11 6,809, ,468 Data sets obtained are quite representative of Turkey s situation in electricity production. Data collected regarding the thermal plants operational performance corresponds to 71% of Turkey s total thermal electricity production at year 2001; while, regarding the renewables, this ratio is 82.6%. This data is later normalized to increase the effectiveness of the DEA. 4. THE DATA ENVELOPMENT ANALYSIS Data Envelopment Analysis (DEA) is a linear programming (L.P.) based technique that provides an objective assessment of the relative efficiency of similar organizational units. In their original paper Charnes et al (1978) introduced the generic term Decision Making Units (DMU) to describe the collection of firms, departments, or divisions which have multiple incommensurate inputs and outputs and which are being assessed for efficiency. Sinse then it has been successfully deployed in many different sectors to assess and compare the efficiency of DMUs [Banker et al (1992), Dyson et al (1987), Golany et al (1994), Korhonen et al (2003)]. The DEA models deployed in this study are summarized below

5 4.1 The CCR Efficiency Model: This model is also called the Global Efficiency model. The main assumption behind it is constant returns to scale, under which the production possibility set is formed without any scale effect. The L.P. model deployed to generate the CCR efficiency factors of the DMUs considered is as follows [Charnes et al (1978)]. Model 4.1: The CCR Model (to be solved for each DMU k 0 ) n max θ CCR (k 0 ) = u j y jk0 j=1 s.t. m v i i=1 m x ik 0 =1 n v i x ik + u j y jk 0 k = 1,, K i=1 j=1 u j 0 j = 1,, n v j 0 i = 1,, m where; u j = the weight for output j; K = the number of DMUs; v i = the weight for input i; y jk = the amount of output j of DMU k; m = the number of inputs; x ik = the amount of input i of DMU k. n = the number of outputs; Plant Table 2. Operational Performance Data Set for Thermal Power Plants Fuel Cost (million TL) Utilization Percentage Production (MWh) CO (Tone) Thermal Eff. (%) Environmental Cost (1000 $) Public 1 32,125, ,673,673 1, ,063 Public 2 57,670, ,724, ,520 Public 3 51,872, ,393, ,119 Public 4 34,623, ,412, ,066 Public 5 32,125, ,637, ,866 Public 6 36,719, ,308, ,724 Public 7 40,653, ,373,340 1, ,066 Public 8 7,153, , ,191 Public 9 120,779, ,016,160 1, ,707 Public 10 46,338, ,534, ,873 Public 11 35,541, ,127, ,781 Public 12 25,593, ,005, ,219 Public ,672, , ,216 Public ,168, ,948,294 1, ,196 Public ,008, ,137, ,349 Public ,576, ,486,842 1, ,500 Public ,724, ,473,877 1, ,198 Public 18 9,871, , ,299 Private 1 241,200, ,600, ,032 Private 2 482,400, ,200,000 1, ,963 Private 3 482,400, ,800,000 1, ,843 Private 4 49,965, , Private 5 40,311, ,

6 Private 6 39,518, , Private 7 3,244, , Private 8 5,904, , Private 9 2,887, , Private 10 4,908, , Private 11 5,045, , The BCC Efficiency Model: This is also called the Local Efficiency (LE) model. The main assumption behind it is variable returns to scale, under which the production possibility set is the convex combinations of the observed units. The L.P. model deployed to generate BCC efficiency factors of the DMUs is as follows [Banker et al (1984)]. Model 4.2: The BCC Model (to be solved for each DMU k 0 ) n max θ BBC (k 0 ) = u j y jk0 u(k 0 ) j=1 s.t. m v i i=1 m x ik 0 =1 v i x ik + u j y jk u(k 0 ) 0 k = 1,, K i=1 u j 0 v j 0 n j=1 j = 1,, n i = 1,, m The inefficiency that a DMU might exhibit may have different causes: whether it is caused by the inefficient operation of the DMU itself or by the disadvantageous conditions under which the DMU is operating, is an important issue to be clarified [Boussofiane et al (1991), Mahgary et al (1995)]. In this regard, comparisons of the CCR and BCC efficiency scores deserve attention. The CCR model assumes a radial expansion and reduction of all observed DMUs (and their nonnegative combinations are possible); while the BCC model only accepts the convex combinations of the DMUs as the production possibility set. If a DMU is fully (100%) efficient in both the CRR and BCC scores, it is operating at the most productive scale size. If a DMU has the full BCC score, but a low CCR score, then it is locally efficient but not globally efficient due to its scale size. Thus, it is reasonable to characterize the scale efficiency of a DMU by the ratio of the two scores. So, scale efficiency is defined as, SE = θ CCR /θ BCC, where θ CCR and θ BBC are the CCR and BCC scores of a DMU, respectively. By definition, SE cannot be greater then one, while the global efficiency (GE), the local efficiency (LE) and the scale efficiency concepts are mathematically related as, GE = LE - SE This decomposition, depicts the sources of inefficiency, i.e., whether it is caused by inefficient operation (LE) or by disadvantageous conditions caused by scale (SE) or both. 4.3 The AR-I-C Efficiency Model It is a version of the CCR efficiency, where weights of all input and output factors are constrained to be in some predetermined regions ( Assurance Regions ). The L.P. model deployed to generate AR-I-C efficiencies of DMUs considered is as follows [Dyson, (1988)]. Model 4.3: The AR-I-C Model (to be solved for each DMU k 0 )

7 max θ ARC (k 0 ) = u j y jk0 s.t. m v i i=1 m x ik 0 =1 n j=1 v i x ik + u j y jk 0 k = 1,, K i=1 n j=1 u i L j u j u i B j v j l i v i v j b i u j 0 v j 0 i = 1,, m; j = 1,, n i = 1,, m; j = 1,, n j = 1,, n i = 1,, m where; Lj = lower bound for the ratio of the weight of output factor j to output factor 1; B j = upper bound for the ratio of the weight of output factor j to output factor 1; l i = lower bound for the ratio of the weight of input factor i to input factor 1; b i = upper bound for the ratio of the weight of input factor i to input factor The AR-I-V Efficiency Model This is a version of the BCC efficiency, where weights of all input and output factors are constrained to be in some predetermined regions ( Assurance Regions ). The L.P. model deployed to generate AR-I-V efficiency factors of the decision making units considered is as follows (Tone 1999). Model 4.4: The AR-I-V Model (to be solved for each DMU k 0 ) n max θ ARV (k 0 ) = u j y jk0 u(k 0 ) j=1 s.t. m v i i=1 x ik 0 =1 m v i x ik + u j y jk u(k 0 ) 0 k = 1,, K i=1 n j=1 u i L j u j u i B j v j l i v i v j b i u j 0 v j 0 i = 1,, m; j = 1,, n i = 1,, m; j = 1,, n j = 1,, n i = 1,, m The Assurance Region Models (namely Model 4.3 and 4.4) require the designation of upper and lower bounds for factor weights (that is, the maximum and minimum possible realistic weights each individual factor may acquire). This is actually a very reasonable requirement, since in practice it is neither realistic nor acceptable to have a generally inefficient plant acquiring a high efficiency rating through the assignment of extremely high (or low) weights to some of its input or output factors. In the assurance region models, such

8 bounds for each individual input (output) factor weight are declared as the minimum and maximum possible ratios of all pairs of input (output) factor weights. In this study, after interviews with various experts, the values of Assurance Region models upper and lower bounds are subjectively set such that, the ratio of any two input (output) factor is at least 0.1 and at most 10. This way, in the process of maximizing a particular plant s efficiency score, for that specific plant, the optimization models may assign individualized values to input and output factor weights such that any one input (output) factor may receive a value up to 10 times of any other input (output) factor weight. Accordingly, the tendency of assigning an extremely high (or low) weight to overemphasize (or underemphasize) certain factors (and thus artificially increasing the efficiency score of an inefficient plant) is restrained. Performance evaluation of considered power plants is done through the DEA models developed. The CCR, the BCC, the AR-I-C, the AR-I-V efficiency scores of all plants are determined; then scale efficiencies are calculated based on these scores. Throughout this study, the software package developed by Saitech Inc. is used for the basic DEA computations (primarily the imbedded L.P. optimization procedures). 5. RELATIONSHIPS BETWEEN EFFICIENCY AND INPUT/OUTPUT FACTORS The DEA results provide the opportunity for the investigation of relationships between plant efficiency values (global and/or scale efficiency) and some of the important input and output factors investigated. In all cases, the curve fittings obtained are based on the least squares methodology (i.e. minimization of the total squared deviation between the curve and data points) and Microsoft Excel is used for the computation of the related parameters. In each case, several possible functional relationships are tried and the one with highest R 2 value is presented. First the case of renewable power plants is considered, based on the AR-I-V model scale efficiency results for investment performance. Figure 1 shows the minimum deviation curve fitting to investment cost/scale efficiency realizations (data points) regarding renewable plants. Figure 2 shows a similar curve fit to installed power versus scale efficiency realizations of renewable plants. Relationships (the curve fits) obtained are weak (R 2 values are hardly satisfying); however, a decreasing exponential relationship seems quite clear

9 Investment vs Scale Efficiency 1,2000 1,0000 0,8000 Scale Efficiency 0,6000 0,4000 0,2000 y = 0,1468x -0,2303 R 2 = 0,3294 0,0000 0,0000 0,2000 0,4000 0,6000 0,8000 1,0000 1,2000 Investment Figure 1. Scale efficiency versus investment cost in renewable plants investment performance Next, operational performance scale efficiency of renewable power plants is considered. Initially, no dominating trend was visible; however, it is observed that it is the very small size plants disturbing the general trend. So, to get a general trend in larger sizes, very small plants (i.e. producing less than 10% of the maximum producer) are removed from the data set and in the resulting curve fit a weak (R 2 around 0.4) but clear exponentially decreasing relationship is obtained. Then, to see the effect of larger plant size, only 10 largest renewable power plants are taken into account: Figure 3 displays the resulting stronger relationship between production and scale efficiency. Again, a clear decreasing exponential relationship emerges for the larger plant data set; in this case, R 2 value is very much satisfying

10 Power vs Scale Efficiency 1,2000 1,0000 0,8000 Scale Efficiency 0,6000 0,4000 y = 0,1262x -0,2833 R 2 = 0,4421 0,2000 0,0000 0,0000 0,2000 0,4000 0,6000 0,8000 1,0000 1,2000 Power Figure 2. Scale efficiency versus installed power in renewable plants investment performance Net Production vs AR Scale Eff of First Ten Most Producing Plants 1,0 0,9 0,8 0,7 Scale Efficiency 0,6 0,5 0,4 0,3 y = 0,295x -0,3415 R 2 = 0,8649 0,2 0,1 0,0 0,0 0,2 0,4 0,6 0,8 1,0 1,2 Production Figure 3. Scale Efficiency versus production in large renewable plants operational performance Another interesting relationship detected is the one between total cost and AR-I-C efficiency. Figure 4 shows the curve fit to efficiency score versus total cost regarding renewable power plants. The relationship obtained is weak (R 2 =0.3), however a decreasing exponential relationship seems clear. Three outliers that are far from the fitted curve are the largest HEPP in Turkey. When they are removed from data set, R 2 value rises up to

11 Cost vs AR-I-C Scores 1,0 0,9 0,8 0,7 0,6 Cost 0,5 0,4 0,3 0,2 y = 0,0139x -0,8611 R 2 = 0,3028 0,1 0,0 0,0 0,1 0,2 0,3 0,4 0,5 0,6 0,7 0,8 0,9 1,0 Efficiency Score Figure 4. AR-I-C efficiency versus cost in renewable plants operational performance Next, investment performance of thermal power plants is analyzed. First, the relationship between AR-I-C efficiency and utilization is considered. Figure 5 displays the detected linear relationship (which has an acceptable R 2 value of 0.71). The slope of this relationship reflects the positive effect of utilization on AR-I-C efficiency. Another relationship highlighted is the one between the AR-I-V efficiency and utilization. The resulting curve fit is similar to the one in Figure 5; however it has a lower R 2 value (0.68). AR-I-C vs Utilization Percentage 1,2000 1,0000 AR-I-C Efficiency Score 0,8000 0,6000 0,4000 y = 1,4014x - 0,4948 R 2 = 0,7128 0,2000 0,0000 0,0000 0,2000 0,4000 0,6000 0,8000 1,0000 1,2000 Utilization Percentage Figure 5: AR-I-C efficiency versus utilization in thermal plants investment performance Next, the relationship between the AR-I-V efficiency and construction time of thermal plants is investigated. It is displayed in Figure 6; the relationship is not strong (R 2 =0,65), but it shows a decreasing exponential trend (as construction time increases, efficiency decreases). This highlights the importance of construction time in investment performance. Then, a similar relationship between AR-I-V efficiency and construction time is identified (R 2 = 0.63)

12 AR-I-V vs Construction Time 1,4000 1,2000 AR-I-V Efficiency Score 1,0000 0,8000 0,6000 0,4000 y = 0,1867x -0,7831 R 2 = 0,6573 0,2000 0,0000 0,0000 0,2000 0,4000 0,6000 0,8000 1,0000 1,2000 Construction Time Figure 6: AR-I-V efficiency versus construction time in thermal plants investment performance Next, scale efficiency in the operational performance of thermal plants is considered. Figure 7 displays the detected strong relationship between scale efficiency and production. This implies that larger plant size decreases efficiency. For further analysis, the data set is divided into two, as gas and coal. Resulting curve fits also support the relationship highlighted in Figure 7. These findings show that small and middle size plants should be preferred when planning new generation investments since they are easier to manage and control AR Scale Efficiency vs Production y = x R 2 = Production Figure 7: Scale efficiency versus production in thermal plants operational performance Effects of production on AR-I-C efficiency are also investigated. Figure 8 shows the curve fitting to production versus AR-I-C efficiency realizations regarding thermal power plants. The decreasing exponential relationship obtained is quite strong (R 2 =0,86), which again confirms the strong influence of production on efficiency of thermal power plants operational performance

13 1,2000 AR-I-C efficiency score vs Production 1,0000 AR-I-C Efficiency Score 0,8000 0,6000 0,4000 0,2000 y = 0,0315x -0,5535 R 2 = 0,8675 0,0000 0,0000 0,2000 0,4000 0,6000 0,8000 1,0000 1,2000 Production Figure 8: AR-I-C efficiency versus production in thermal plants operational performance Another important factor investigated is effects on the environment from an operational performance viewpoint. Figure 9 and 10 show the curve fits regarding environmental cost versus AR-I-C efficiency and environmental cost versus scale efficiency, respectively. Relationships obtained are not strong (R 2 values are quite low); however, decreasing exponential relationships are quite clear. Thus, it can be said that reducing negative environmental effects, will favorably effect AR-I-C or scale efficiency realizations AR-I-C efficiency vs Environmental Cost y = x R 2 = Environmental Cost Figure 9: AR-I-C efficiency versus environmental cost in thermal plants operational performance 6. EFFICIENCY COMPARISONS AMONG TYPES OF POWER PLANTS Regarding HEPP, comparisons are made between channel and reservoir type plants. Figure 11 shows the results obtained (the values displayed are the averages). The main difference is the AR-I-V score. Channel types have higher AR-I-V efficiency values, which indicate that they are managed more effectively. In spite of this, reservoir type has slightly higher AR-I-C score

14 The reason for this difference may be the higher production potential and thus the higher scale efficiency of reservoir types. 1,2000 AR Scale efficiency vs Environmental Cost 1,0000 Scale Efficiency 0,8000 0,6000 0,4000 0,2000 y = x R 2 = ,0000 0,0000 0,2000 0,4000 0,6000 0,8000 1,0000 1,2000 Environmental Cost Figure 10: Scale efficiency versus environmental cost in thermal plants operational performance Comparison of Channel vs Reservoir Types 0,70000 Channel Type With Reservoir 0, ,50000 Normalized Values 0, , , , ,00000 AR-I-C Score Production Total Cost Working Hour AR-I-V Score Scale Eff With Reservoir Channel Type Figure 11: Comparison of channel type and reservoir type HEPP Next, investment performance of public and private sector thermal power plants are compared; they are displayed in Figure 12. This figure indicates that, on average, the private sector plants have higher AR-I-C and AR-I-V efficiency scores than the public sector plants. On the other hand, scale efficiency scores of the two sectors plants are close to each other. Another observation is that, on average, investment cost of public sector plants is higher than that of the private sector plants (indicating the more expensive projects of the public sector). Furthermore, construction time of the public sector plants is also higher (this points to longer project completion times and longer lead times); moreover, public sector owns larger

15 size plants. Another interesting point is that utilization percentage of the public sector plants is less than that of the private sector. Public versus Private Natural Gas Plants 1,0000 Public Natural Private Natural Normalized Value 0,9000 0,8000 0,7000 0,6000 0, , ,3000 0,2000 0,1000 0,0000 AR-I-C Score AR-I-V Score Scale Efficiency Investment Cost Construction Time Figure 12: Comparison of public and private sector thermal plants by investment performance For additional insight, the set of thermal power plants is decomposed according to input source (coal and natural gas). The comparison of these two subsets is displayed in Figure 13. Based on this figure, coal plants seem to be at a clear disadvantage with respect to natural gas plants: On average, natural gas has higher AR-I-C and AR-I-V efficiency scores relative to coal. Furthermore, investment cost and construction time values of coal plants are higher than natural gas ones; utilization percentages also show the dominancy of natural gas. Power Utilization Percentage Private Natural Public Natural Figure 13: Comparison of private and public natural gas plants by investment performance

16 Another interesting comparison is between the private and public natural gas plants. Figure 14 displays a comparison between these two types of power plants. According to this figure, averages of AR-I-C, AR-I-V and AR Scale efficiency values show the public sector to be slightly more efficient than the private sector. Higher investment cost of public sector comes from the larger size plants they operate; accordingly, construction time of public plants is also somewhat higher. An interesting point is that utilization percentage of public gas plants is lower than their privately owned counterparts. This decomposition and analysis hints that the main drawback of the public sector may be due to the coal-fired plants it owns. Better management and adaptation of new technologies to these coal fired plants can improve the situation. In this regard, more research and development activities are needed specifically for the deployment of low quality coal. Coal vs Natural Gas Plants 1,0000 0,9000 0,8000 Coal Natural Normalized Values 0,7000 0,6000 0,5000 0,4000 0,3000 0,2000 0,1000 0,0000 AR-I-C Score AR-I-V Score Scale Efficiency Investment Cost Construction Time Power Utilization Percentage Figure 14: Comparison of coal and natural gas fired plants by investment performance. Coal Natural Public Sec vs Private Sec 1 0,9 0,8 0,7 0,6 0,5 0,4 Normalized Value 0,3 0,2 0,1 Public Sector Private Sector AR-I-C Score AR-I-V Score AR Scale Efficiency Production Utilization Environment 0 Figure 15: Comparison of public and private thermal plants by operational performance

17 Figure 15 displays the operational efficiency comparisons of the public and private thermal power plants. As can be seen from the figure, the private plants have higher AR-I-C, AR-I-V and scale efficiency scores. Especially AR-I-V scores hint at the better managerial skills of the private sector. Average productions of the two subsets are similar (which actually enables a more reliable comparison regarding the other attributes). Utilization percentages show the private sector to be more advantageous. An interesting point is environmental cost incurred by the public sector plants: The public plants have relatively higher environmental cost. The comparison in Figure 16 is based on fuels used. It can be observed that natural gas plants are more environmentally friendly than coal fired ones. The higher AR-I-V efficiency of gas plants is indicative of their easier (or better) management. (they also have higher utilization rates). On the other hand, scale inefficiency is more dominant at coal-fired plants. Natural Gas vs Coal vs Liquid Plants 1,0000 0,9000 0,8000 Normalized Value 0,7000 0,6000 0,5000 0,4000 0,3000 0,2000 Liquid Coal Natural 0,1000 0,0000 AR-I-C Score AR-I-V Score AR Scale Efficiency Production Utilization Environment Figure 16: Thermal power plants comparison based on operational performance Liquid Coal Natural Private versus Public Natural Gas Plants Normalized Value AR-I-C Score AR-I-V Score AR Scale Efficiency Production Utilization Environment Private Natural Public Natural Figure 17: Public and private gas plants comparison based on operational performance

18 Figure 17 displays the comparison of public and private natural gas plants (which is the type having highest efficiency scores). Regarding AR-I-V efficiency, the public rates better. But public s scale inefficiency makes its AR-I-C score worse that that of the private sector. Publicly owned natural gas power plants also have disadvantages on the environment side. 7. CONCLUSION In this study, performance of 65 thermal, hydro and wind power plants, owned by the private and public sectors in Turkey have been analyzed and compared through Data Envelopment Analysis (DEA). Two basic efficiency models, reflecting operational and investment performance, have been defined and pursued. Constant returns to scale, variable returns to scale and assurance region type DEA models are used in the analysis. Various performance comparisons have been conducted and also, relationships between efficiency scores and various input / output factors have been investigated and identified. Interesting conclusions that can be drawn from this study are summarized as follows: Regarding renewable source power plants, (excluding very small size plants) scale efficiency shows an exponentially decreasing trend with respect to plant size (installed power and/or production). Regarding thermal power plants, global and scale efficiencies show a strong exponentially decreasing trend with respect to plant size (production). Regarding thermal power plants, global and scale efficiencies show a linearly increasing trend with respect to utilization percentage. Regarding thermal power plants, global and scale efficiencies show an exponentially decreasing trend with respect to construction time. Without reducing negative environmental effects (costs) of thermal power plants, it would not be possible to reach higher efficiency scores. Regarding thermal power plants investment performance efficiency, the private sector plants perform significantly better than the public sector plants. This is hardly surprising, since public plants have higher investment costs, longer construction times and lower utilization percentages. One main cause of investment performance inefficiency of the public sector is because of the significantly poorer performance of its coal-fired plants. Natural gas fired power plants have higher investment performance efficiency than coal fired plants. Public. gas power plants have slightly higher investment efficiency than private ones. Operational performance efficiency of the public thermal plants is significantly lower than their private counterparts. Operational performance efficiency of coal and liquid fuel fired power plants are considerably lower than that of the natural gas fired ones. Operational performance efficiency of public sector natural gas fired power plants is slightly lower than that of the privately owned ones. Lower operational efficiency of public sector plants is a reflection of the high ratio of coal and liquid fuel fired plants in its portfolio. Wind power plants have the highest efficiency values in operational and investment performance models, which is a strong indication of their high future potential. It should be noted that, regarding evaluations and comparisons involving renewable source plants, fewer conclusions could be drawn. This might be related to the higher performance variation of renewable source plants due to their higher sensitivity to natural factors (like wind, amount of rain fall etc.); thus relationships found are small and insignificant. The significant difference between the utilization percentages of public and private thermal plants (which has an important effect on efficiency realizations) may be related to the

19 associated operating conditions and/or contracts (such as the sales guarantees of the Build- Operate (BO) and Build-Operate-Transfer (BOT) contracts). This issue is not investigated. REFERENCES Bakos, G.C. and Tsagas N.F. (2003), Technoeconomic Assessment of a Hybrid Solar/Wind Installation for Energy Saving, Energy and Buildings, 35, Banker, R.D., Charnes A. and Cooper W.W. (1984), Some Models for Estimating Technical and Scale Inefficiencies in Data Envelopment Analysis, Management Science, 30.9, Banker, R.D., Thrall R.M. (1992), Estimating Most Productive Scale Size Using Data Envelopment Analysis, European Journal of Operational Research, 62, Boussofiane, A., Dyson R.G. and Thanossoulis E. (1991), Applied Data Envelopment Analysis, European Journal of Operational Research, 52, Charnes, A., Cooper W.W. and Rhodes E. (1978), Measuring the Efficiency of Decision Making Units, European Journal of Operational Research, 2, Dyson, R.G. and Thanossoulis E. (1988), Reducing Weight Flexibility in Data Envelopment Analysis, Journal of Operational Research Society, 39.6, Dyson, R.G., Thanassoulis E. and Foster M.J. (1987), Relative Efficiency Assessments Using Data Envelopment Analysis: An Application to Data on Rates Departments, Journal of Operational Research Society, 38.5, ExternE (1995), Externalities of Energy- 2, 3, 6, The European Commission, Luxembourg, ( 2003). Gabriel, S.A., Kydes A.S. and Whitman P. (2001), The National Energy Modeling System: A Large-Scale Energy-Economic Equilibrium Model, Operations Research, 49.1, Golany, B., Roll Y. and Rybak D. (1994), Measuring Efficiencies of Power Plants in Israel by Data Envelopment Analysis, IEEE Transactiıns on Engineering Manag., Hondroyiannis, G., Lolos S., Papapetrou E. (2002), Energy Consumption and Economic Growth: Assessing the Evidence from Greece, Energy Economics, 24, IEA-International Energy Agency (2001), Energy Policies of IEA Countries Turkey 2001 Review, Paris. Korhonen, P.J. and Luptacik M. (2003), Eco-efficiency Analysis of Power Plants: An Extension of Data Envelopment Analysis, European Journal of Operational Research. Kwak, H.Y., Kim D.J., and Jean J.S. (2003), Exergetic and Thermoeconomic Analysis of Power Plants, Energy, 28, Liu, Y., et al (2003), Economic Performance Evaluation Method for Hydroelectric Generating Units, Energy Conversion and Management, 44, Mahgary, S. and Lahdelma R. (1995), Data Envelopment Analysis: Visualizing The Results, European Journal of Operational Research, 85, Pacudan, R., Guzman E. (2002), Impact of Energy Efficiency Policy to Productive Efficiency of Electricity Distribution in the Philippines, Energy Economics, 24, Park, S.U. and Lesourd J.B. (2000), The Efficiency of Conventional Fuel Power Plants in South Korea: A Comparison of Parametric and Non-parametric Approaches, International Journal of Production Economics, 63, Sàez, R.M., Linares P. and Leal J. (1998), Assessment of the Externalities of Biomass Energy and Comparison of Costs with Coal, Biomass and Bioenergy, 14, Tone, K. (1999), On Returns to Scale Under Weight Restrictions in Data Envelopment Analysis, Research Report Series I , U.S. Nat. Grad. Inst. for Policy Studies