A case study to evaluate the productivity changes of the thermal power plants of the Taiwan power company



A Case Study to Evaluate the Productivity Changes

of the Thermal Power Plants of the

Taiwan Power Company

Chen-Fu Chien, Member, IEEE, Wen-Chih Chen, Feng-Yu Lo, and Yi-Chiech Lin

Abstract—This paper developed an approach based on data

en-velopment analysis and Malmquist productivity index to investi-gate the performance of power plants and conducted an empirical study with eight thermal power plants in Taiwan. The analysis re-sults show the productivity improvements, and thus, help Taiwan Power Company to monitor and diagnose changes in the produc-tivity of its thermal power plants. Furthermore, this study also provides specific directions for improvements to increase competi-tiveness in the face of the continuing liberalization of the Taiwanese power generation market.

Index Terms—Data envelopment analysis (DEA), Malmquist

productivity index, productivity change, relative efficiency, state-owned power company.



HIS PAPER presents a case study in which we ap-plied data envelopment analysis (DEA) models to mea-sure the relative efficiencies and employed Malmquist pro-ductivity index to measure propro-ductivity changes for thermal power plants operated by the Taiwan Power Company (TPC). TPC is a state-owned enterprise in Taiwan, and is the only electric utility in charge of power generation, transmission, distribution, and sales. Facing the challenges of power in-dustry liberalization, TPC has devoted itself to establishing a technology–productivity–competitiveness oriented operating strategy to maintain its competitiveness in power industry [1]. Moreover, to increase competition in the power market, the Taiwanese government is adopting the policy of privatizing the TPC and issuing new licenses to Independent Power Providers (IPPs). Consequently, it is crucial for the TPC to increase the productivity of thermal power plants to cope with the negative impacts of liberalization of the electricity market and competi-tion from IPPs [2]. TPC needs effective methods to evaluate and examine changes in the productivity of thermal power plants.

This study was motivated by a real problem, and thus, aimed to fill the gap for evaluating the performance of the thermal power plants operated by TPC. In particular, this study employs Manuscript received September 13, 2002; revised July 14, 2006. This work was supported by the Taiwan Power Company (TPC-531-4830-04) and Na-tional Science Council, Taiwan, R.O.C. (NSC91-2218- E-007-007). Paper no. TEC-00209-02002.

C.-F. Chien and Y.-C. Lin are with the Department of Industrial Engineering and Engineering Management, National Tsing Hua University, Hsinchu 300, Taiwan, R.O.C. (e-mail: cfchien@mx.nthu.edu.tw; ej.lin@auo.com).

W.-C. Chen is with the Department of Industrial Engineering and Manage-ment, National Chiao Tung University, Hsinchu 300, Taiwan, R.O.C. (e-mail: wenchih@faculty.nctu.edu.tw).

F.-Y. Lo is with the Taiwan Power Company, Taipei 100, Taiwan, R.O.C. (e-mail: u246456@taipower.com.tw).

Digital Object Identifier 10.1109/TEC.2006.888026

nonparametric Malmquist productivity index to study the pro-ductivity changes, which can be classified into efficiency and technology aspects, of the plants from 1994 to 1999. In prac-tice, efficiency evaluation, generally, considers multiple input resources and output factors. DEA, first introduced by Charnes

et al. in 1978 [3], is a linear programming approach that is

capa-ble of simultaneously dealing with multiple inputs and multiple outputs, yet which does not require the predetermined weights to the input and output factors. DEA models offer a practica-ble approach for evaluating the relative efficiency of decision-making units (DMUs) in various contexts. However, the relative efficiency and the productivity changes of power plants have seldom been addressed in existing studies.

The rest of this paper is organized as follows. Section II de-scribes the fundamentals of this study including DEA models and Malmquist productivity index and also reviews the related literature. Section III describes an empirical study on the pro-ductivity changes of thermal power plants in Taiwan. Section VI concludes with discussions on the contributions and future re-search directions.


A. Distance Functions and DEA Models

Power generation is a process transforming multiple inputs (resources) to various outputs (services). In this section, we, first, generalize and formalize this transformation process in terms of models, and thus, introduce some relevant techniques based on the generalization for performance evaluation.

Considering x∈ m

+ as the input vector and y∈ n+ as the

output vector, the production technology Γ such as power gen-eration can be defined as follows:

Γ≡ {(x, y) : y can be produced by x}. (1) The boundary of the technology is defined as

∂Γ≡ {(x, y) : (x, βy) /∈ Γ, β > 1}. (2) Shephard [4] defined the output distance function between any specific input–output bundle (x, y) and the boundary ∂Γ as follows:

DO(x, y)≡ inf {α : (x, y/α)∈ Γ} (3) where DO(x, y) can characterize the relative location of (x, y) to Γ. Thus, (x, y)∈ Γ if and only if DO(x, y)≤ 1; (x, y)∈ ∂Γ if and only if DO(x, y) = 1.


Although the production technology Γ is unknown, given a set of observations R ={(x1, y1), (x2, y2), . . . , (x|R|, y|R|)}, where xr∈ m+and yr∈ n+for r∈ R, it can be derived based

on the conditions of convexity, free disposal, inclusion of obser-vations, and minimum extrapolation [5]. The empirical produc-tion technology ˆΓ derived from the observation set can, then, be expressed as a set of linear inequalities in nonnegative variables as follows: ˆ Γ  (x, y) : r∈R xrλr≤ x,  r∈R yrλr≥ y  r∈R λr= 1,λr≥ 0, r ∈ R. (4)

Therefore, the output distance function DO(x, y) for any (x, y) corresponding to ˆΓ can be estimated as follows:

[DO(x, y)]−1= max α s.t.  r∈R xirλr≤ xi, i = 1, . . . , n  r∈R yjrλr≥ αyj, j = 1, . . . , m  r∈R λr= 1 λr≥ 0, r ∈ R. (5) If (x, y)∈ R, i.e., (x, y)∈ Γ, the output distance functions can also be interpreted as the output-based Debreu–Farrell tech-nical efficiency (TE) measures [6], [7], the relative efficiency compared against the peer group R. Combining the concepts of TE and empirical production technology leads to one of the well-known DEA models, i.e., the Banker–Charnes–Cooper (BCC) model [5] as follows: max θ s.t.  r∈R xirλr≤ xi, i = 1, . . . , n  r∈R yjrλr≥ θyj, j = 1, . . . , m  r∈R λr= 1 λr≥ 0, r ∈ R. (6) The optimal value of (6), θ∗, is related to the output-oriented TE for DMU (x, y)∈ R. For example, if θ∗= 1.2, it means that, compared to the given peer group, this specific DMU (x, y) can increase all its outputs to 120% while maintaining same input usage since a feasible production plan that uses no more than x to produce no less than 1.2ycan be derived from the observed DMUs.

Assuming constant returns to scale (CRS), (6) can be rewrit-ten as another popular DEA model, i.e., the Charnes–Cooper–

Rhodes (CCR) model [3] as follows: max φ s.t.  r∈R xirλr≤ xi, i = 1, . . . , n  r∈R yjrλr≥ φyj, j = 1, . . . , m λr≥ 0, r ∈ R. (7) Other DEA variants and more comprehensive reviews can be found, for example, in [8]–[11].

Banker et al. [5] referred to 1/φ∗ as the overall efficiency, 1/θ∗as the pure TE, and the ratio θ∗/φ∗as the scale efficiency (SE). TE determines the operational efficiency. Moreover, a DMU with TE less than one, i.e., θ∗> 1, indicates that the

current operation is inefficient and output should be able to be increased while maintaining the same input level. SE measures the difference from the optimum production scale, i.e., the most productive scale size (MPSS). The smaller the SE, the worse the production scale or further difference from MPSS.

The optimal solutions of (7) also guide the direction of im-provements in relation to scale. r∈Rλ∗r< 1 indicates that the DMU operates at increasing returns to scale (IRS) region, i.e., a proportional increase of all input levels would produce a more than proportional increase in output levels. A DMU with 

r∈Rλ∗r> 1 operates at decreasing returns to scale (DRS) region, in which a proportional increase of all input levels pro-duces a less than proportional increase in output levels. If a DMU is operating at the MPSS, it is scale efficient, and thus, presents CRS. That is,r∈Rλ∗r= 1.

Alternatively, a scale-inefficient DMU presents DRS as it ex-ceeds the MPSS and presents IRS as it is lesser than the most productive size. This implies that resources may be transferred from DMUs operating at DRS to those at IRS to increase pro-ductivity at both sets of DMUs. That is, using DEA models can specify the directions of resource reallocations to improve SE as well as overall efficiency.

B. Malmquist Productivity Index

Malmquist index is used for bilateral comparisons of the pro-ductivity change over time or the propro-ductivity difference for any different predefined peer groups [12]. Fare et al. proposed a DEA-like nonparametric Malmquist index that can be decom-posed into technical change and efficiency change [13], [14].

Let Γb and Γtrepresent the production technologies during the base period b and another period t, respectively. Hereafter, the superscripts represent the production technology and/or data from the corresponding periods b or t. Considering (xb, yb)∈ Γb and (xt, yt)∈ Γt, two output distance functions can be defined as follows:

DbO(xb, yb)≡ inf {α : (xb, yb/α)∈ Γb} (8)

DOt(xt, yt)≡ inf {α : (xt, yt/α)∈ Γt} (9) where Db

O(xb, yb) and DOt(xt, yt) measure the distance from (xb, yb) and (xt, yt) to the corresponding technology


boundary ∂Γb and ∂Γt. It is clear that Db

O(xb, yb)≤ 1 and


O(xt, yt)≤ 1. The projected points are obtained as follows:xb, ˜yb)  xb, y b Db O(xb, yb)  ∈ ∂Γb and (˜xt, ˜yt)  xt, y t Dt O(xt, yt)  ∈ ∂Γt.

They are on the technology boundaries, and so, both are techni-cally efficient. The distance function from (˜xb, ˜yb) to the bound-ary of Γtis, thus, the boundary shift measure from ∂Γbto ∂Γt with regard to (xb, yb), where the input/output mix remains unchanged, and can be expressed as follows:

DtOxb, ˜yb)≡ inf {α : (˜xb, ˜yb/α)∈ Γt}. (10) Similarly, the backward distance from (˜xt, ˜yt) to ∂Γb, i.e., the boundary shift from ∂Γtto ∂Γb, is defined as follows:

DbOxt, ˜yt)≡ inf {α : (˜xt, ˜yt/α)∈ Γb}. (11) It should be noted that DtOxb, ˜yb) and DbOxt, ˜yt) do not nec-essarily need to be equal to or less than one. The values being greater than one indicates that the base boundary is outside the reference technology, and thus, the base technology outperforms the reference technology. In particular, Db

Oxt, ˜yt) > 1 indi-cates that (˜xt, ˜yt) /∈ Γb. If technological improvement occurs from b to t, Dt

Oxb, ˜yb) < 1, yet DObxt, ˜yt) > 1. Moreover, (5) can be used as a practical computation procedure for all distance functions discussed here with some minor changes for the peer group R and the reference basis (x, y).

The geometric mean of Dt

Oxb, ˜yb) and DObxt, ˜yt)−1is the technical change that measures the shift in frontier technology (SIT) between the two time periods b and t [13]. The technical change can, thus, be written as follows:

Technical change =  Db Oxt, ˜yt)DOtxb, ˜yb)−1 =  Db O(xt, yt) Dt O(yt, yt) Db O(xb, yb) Dt O(xb, yb) . (12) If technical change exceeds one, it indicates that the production technology or capability progresses from period b to t. Mean-while, if the SIT score is lesser than one, it indicates that the production technology is depressive between the two time peri-ods.

Fare et al. [13] defined the efficiency change between b and t as follows: Efficiency change = D t O(xt, yt) Db O(xb, yb) . (13) Efficiency change measures the degree of catching-up in ef-ficiency (CIE) from period b to t. A CIE score exceeding one indicates an improvement in efficiency, while a CIE score be-low one implies a decrease in efficiency during these two time periods.

The product of technical change (or SIT) and efficiency change (or CIE) is the overall productivity change between

b and t, which is conceptually consistent with the Malmquist

index [12]. A score of Malmquist index exceeding one indi-cates overall productivity improvement during these two time periods.

Both the SIT and CIE scores can be evaluated for under-standing the productivity change of the DMU during a specific time period. For example, Chang et al. [15] applied the DEA model and the Malmquist productivity approach to evaluate the relative changes of 23 Taiwan administrative regions between 1983 and 1990. Moreover, Chen and Yeh [16] evaluated the rel-ative efficiency of 34 commercial banks and used the Malmquist productivity index to investigate changes in their productivity between 1995 and 1996.

C. DEA Studies

There are a number of DEA studies on power systems and power industry, while DEA has been successfully applied in different domains. Extensive reviews and additional applications can be found in [10] and [11].

For studies on power industry, Fare et al. [17] applied DEA to evaluate the relative efficiency of electric utilities regulated by Illinois Commerce Commission, in which an output (net generation) and three inputs (fuel, labor, and capital) were con-sidered. Charnes et al. [18] applied the DEA model to evaluate the management efficiency of regulated electricity cooperatives in Texas, in which three outputs (net margin, total kilowatthour sales, and total revenue from sales of electricity) and 11 in-puts (operations expenses, maintenance expenses, consumer ac-counts expenses, administrative and general expenses, miles per consumer, line loss, average outage hours per customer, per-centage system unload, total plant, salaries, and inventory) were considered. Miliotis [19] measured the efficiency of 45 electric-ity distribution districts of the Greek Public Power Corporation in different settings and used four different sets of input and output factors including the number of customers served, net-work length, capacity of installed transformation points, general expenses, administrative labor hours, technical labor hours, sup-plied energy, and served area for different analyses. In addition, the authors applied DEA models to evaluate the relative oper-ating efficiency of 22 electricity distribution districts of TPC in Taiwan, and identified specific improvement directions for the inefficient districts [20], in which two outputs (total number of customers and energy supplied) and five inputs (employ-ment expenditure, general expenditure, total assets, distribution network, and transformer capacity) were considered. In addi-tion, we also investigated the reorganization of the districts and proposed organizational changes for improving their operating efficiency. Therefore, we conducted further study to improve the operational efficiency of the poorest district [21] based on DEA analysis. Also, we developed effective data mining methodolo-gies to help TPC for fault location on power distribution feeder to reduce the impact of electricity shortage, and thus, improve operational efficiency [22]–[24]. Pahwa et al. [25] conducted a similar study for evaluating the relative efficiency of 50 elec-tricity distribution centers in the United States.


For studies on power plants, Athanassopoulos et al. [26] de-veloped the data envelopment scenario analysis for setting tar-gets to electricity generating plants in the United Kingdom, in which four outputs (electricity produced, plant availability, accidents incurred, and pollution generated) and three inputs (fuel, controllable costs, and capital expenditure) were consid-ered in the DEA models. Sueyoshi [27] explored a marginal cost-based pricing measurement using the cost-based DEA ap-proach to examine the tariff structure of nine electric power companies in Japan, in which the output of 11 electricity sales (residential services, commercial services, and other services) and three inputs (labor price, capital price, and materials price) were considered. Park and Lesourd [28] evaluated the operating efficiency of 64 conventional fuel power plants in South Korea and considered an output (net electrical energy) and three in-puts (fuel consumption, installed power, and labor). Sueyoshi and Goto [29] proposed a slack-adjusted DEA model to ex-amine the performance of Japanese electric power generation companies from 1984 to 1993 and considered total generation as the output and used three inputs (capacity, total fuel con-sumption, and total employees). Raczka [30] evaluated the per-formance of 41 thermo-electric power plants in Poland, in which a single output (heat production) and three inputs (labor, fuel, and air pollution penalty) were considered. Finally, Cook and Green [31] developed a two-stage hierarchical model to evalu-ate a set of power plants and the individual power generating units, in which three inputs (forced derating, maintenance ex-penditure, and occupied hours) and two outputs (full capacity operating hours and number of outages) were considered for the analysis.

Furthermore, an increasing number of studies have addressed the growing concern about environmental issues by including pollutants such as SO2emissions in the analysis. In particular,

Golany et al. [32] evaluated the operating efficiency of power plants in the Israel Electric Corporation, in which four outputs (generated power, operational availability, deviation from opera-tional parameters, and SO2emissions) and three inputs (installed

capacity, fuel consumption, and manpower) were considered. Lee et al. [33] studied the Korean electric power industry using data from 1990 to 1995, in which three inputs (labor, capacity, and fuel), one output (annual power generation), and three un-desired outputs (emissions of SOx, NOx, and total suspended particulates) were studied. Fare et al. [34] compared the TE of 209 electric utilities before (in 1993) and after (in 1997) the implementation of the legislation to control acid rain. Fare et al. considered three inputs (labor, capacity, and fuel), one positive output of annual power generation, and one undesired output of SO2emissions and proposed a method of measuring shadow

prices of SO2and the output elasticity of substitution between

SO2and electricity.


This section details an empirical study investigating the pro-ductivity changes of the TPC thermal power plants from 1994 to 1999. Following Golany and Roll [35], this empirical study involves three major tasks including problem structuring,

deter-mination of input and output factors for measuring the relative efficiency of the selected DMUs, and discussion of the DEA results.

A. Problem Structuring

TPC has eight thermal power plants, namely HSIEHHO,


TUNGHSIAO, and PENGHU. The total power generation of these eight plants exceeds 70% of the total energy generated in Taiwan [36]. On one hand, the eight considered plants using the same inputs to produce the same output factors belong to a homogeneous group for evaluation. On the other hand, their scales are very different in terms of number of employees, installed power generation capacity, production costs, and energy generation. That is, the power plants utilize the same input resources at various levels to generate various levels of energy output. It is difficult for TPC to predetermine an effective set of weights to consider multiple input and output factors for evaluating their operation efficiencies and productivity changes meaningfully. Therefore, DEA offers an appropriate method of filling in this gap and evaluating the relative efficiencies of TPC power plants.

B. Input and Output Factors

According to the guidelines in the TPC Responsibility Cen-ter System [37] and Profitability CenCen-ter System [38], this study uses total energy generated as the output factor and total in-stalled capacity (MW), total number of employees, and total production cost as input factors in the DEA models. The total energy generated is the major output since the function of the plants is to supply electricity to meet demand. The installed power generation capacity is a fundamental input factor that differentiates plant productivity. Meanwhile, the total number of employees is an important input, and in fact, personnel cost is also a critical input factor in state-owned enterprises. Production cost, which includes operating expenditures, fuel expenditures, and maintenance expenditures, is the input factor that covers the cost of supporting and maintaining plant operations. The value is derived from annual reports and adjusted by the whole-sale price index (WPI), similar to the well-recognized producer price index (PPI), to constant dollars.

Table I lists the annual data of eight plants used in this study from 1994 to 1999. Using the annual data can reduce the influ-ence of seasonal effects. Moreover, considering six time periods can effectively evaluate the productivity change of the plants along the time.

C. DEA Results for TPC Power Plants

Table II summaries the DEA results in which the eight plants in the same peer group are compared over a six-year period. It is found that all DMUs except for TAICHUNG and HSINTA (1998 and 1999) are found to be in the IRS region. Notably, small plants like PENGHU perform extremely poorly in terms of SE. For those in the IRS region, increasing their production scale offers one way of improving operational efficiency. For




example, HSINTA moved from the IRS region to MPSS after increasing its capacity in 1997 and 1998. The plants indicated as in the optimum production scales, such as TAICHUNG and HSINTA, all have the largest installed capacity. This shows that larger power plants are more advantageous—at least up to the scale of TAICHUNG.

TAICHUNG plant has the largest scale than do other plants, and is classified as MPSS each year. From 1994 to 1999, this plant has increased its resource usages including total installed electricity generating capacity from 2480 kMW to 4680 kMW, while increasing total employees from 601 to 782,

and total production costs from around NT$ 10 702 million to NT$ 18 494 million. Furthermore, TAICHUNG plant has also increased the total generation energy from 15 634 561 million kWh in 1994 to 33 039 905 million kWh in 1999.

On the other hand, PENGHU plant is located on a small island off Taiwan, and is smaller than other plants. From 1994 to 1999, PENGHU plant increased only its total installed electricity generating capacity from 49 kMW to 74 kMW. Table II shows that PENGHU, generally, exhibits high TE scores, indicating that it has good operational management. However, the SE scores of PENGHU plant are only around 0.5,




indicating that the production is not in an inappropriate produc-tion scale, and this aspect deserves further study.

Notably, production technology might have changed during the six-year time frame. Therefore, this study also applies the Malmquist output-based productivity index to investigate the productivity changes of the eight TPC plants from 1994 to 1999 using the panel data listed in Table I.

D. Productivity Changes of the Power Plants

Table III summarizes the results of the overall productivity changes via Malmquist index, which can be further decomposed into technical changes and efficiency changes as addressed ear-lier. The calculations are based on (5), (12), and (13), in which

b = 1994 and t = 1995–1999. That is, 1994 is set as the base

pe-riod to be the reference point for observing the annual changes. As shown in Table III, the overall productivity indices increase from 1994 to 1999, except for HSINTA and LINKUO, for which it decreased slightly. The average overall productivity index was 1.134 in 1999. These indicate the improvements in productivity, and the average productivity in 1999 was around 13.4% better than that in 1994.

Examining the efficiency change (CIE), PENGHU and TAICHUNG remain constant (i.e., CIE score equals one) over the six-year study period. After a significant raise in 1995, the CIE scores of SHENAO remain the same. This trend indicates that PENGHU, TAICHUNG, and SHENAO can maintain the same position relative to others in terms of operational




efficiency. That is, these three plants are able to perform better during all periods. However, the CIE scores of another three plants HSIEHHO, HSINTA, and LINKOU were observed to be continuously decreasing since 1997. This implies that these three plants operate inefficiently compared to the other five dur-ing these time periods, i.e., their own performance is deteriorat-ing, while the performance of other plants is improving. Further checking the technical change index (SIT) of HSIEHHO, HSINTA, and LINKOU reveals increased scores that exceeded one. This phenomenon implies that the production technology is, generally, progressive but the operations of these three remain at a similar level. That is, all the subject power plants improved but the pace of improvement at these three stations lagged the rest of the group. Compared to the data in Table I, the power generated by HSIEHHO and HSINTA has changed little since 1997, yet the capacity of HSINTA increased from 3454 kMW in 1997 to 4626 kMW in 1999. This shows that HSINTA has failed to achieve the expected returns on its capacity increase.

Moreover, as illustrated in Table II, HSINTA followed an effective direction that its production scale has been increas-ing and it was, thus, moved from the IRS region to the MPSS. However, it demonstrates slight gains in terms of technology but significant losses in terms of operating efficiency. Similarly, LINKUO increased its capacity from 785 kMW in 1997 to

1085 kMW in 1998. However, after 1997, its technology con-tinued improving while its efficiency reduced. This phenomenon might indicate that HSINTA and LINKUO do not appropriately adjust their short-term operations to incorporate the increase in capacity. Above observations demonstrate the importance of careful managerial adjustment of operations when the produc-tion scale is changed. Short-term operaproduc-tional efficiency is not only the key to successful productivity improvement without long-term structural changes such as scale, but is also the key determinant of the success of any strategic structural change in an organization.

In Taiwan, the power industry has undergone major changes owing to the liberalization of the electricity market and the privatization of TPC. In 1995, the government began to lib-eralize power generation (mostly thermal power plants). For example, the Mailiao thermal power plant owned by Taiwan’s largest private enterprise installed three power generation units that started commercial operations on June 1, 1999, September 9, 1999, and September 23, 2000, respectively [36]. These new IPPs are, generally, considered to be efficient since they apply new technology, in contrast to the mostly old ther-mal power plants of TPC. Consequently, TPC has also taken actions to maintain and renovate its plants. Facing increasing challenges and decreasing government support, TPC should


strive to improve the operating efficiencies of its plants to rise to the challenges from new IPPs in the deregulated power mar-ket in Taiwan. The proposed approach can, effectively, assist the decision makers to determine which new proposals should be funded, which existing projects should be continued, and what levels of each resource are needed for the thermal power plants. TPC decision makers should also justify their decisions and communicate those decisions to others, including the union, legislators, and government.


Using Malmquist productivity index and its decompositions, this study evaluates productivity changes for eight TPC thermal power plants in Taiwan, from 1994 to 1999. This method can be adopted to monitor and diagnose productivity changes result-ing from management decisions and the effectiveness of their implementation.

The overall productivity changes of the eight plants showed a slight progressive development. The analytical results provide decision makers with useful information regarding specific ar-eas to be considered in improving the plant efficiency. This study finds significant effects of economic scale, i.e., larger production scale is better for TPC thermal power plants. However, while increasing the production scale that leads to progressive produc-tion technologies, greater attenproduc-tion should be paid to short-term managerial effort and operational adjustments. Without incor-porating proper and effective short-term management, relative operational efficiency will decrease, damaging the total produc-tivity improvement even when the long-term strategic direction is correct.

Indeed, most inefficient plants suffered from technical inef-ficiency rather than scale inefinef-ficiency, despite TPC attempting to increase the power generation capacity of its thermal power plants to meet the increasing energy demands in Taiwan. Thus, increased efforts should be made to improve operations and re-source utilization. To maintain its competitive advantage in the face of the future liberalization of the electricity market and privatization, TPC needs to pay more attention not only to eval-uating the relative efficiency and efficiency changes of thermal power plants but also to accomplishing the three major tasks (i.e., improving generation technology, enhancing productivity, and strengthening competitiveness) in the Competitiveness En-hancement Plan [36]. However, since the IPP plants data are confidential, this study could not directly compare the perfor-mance of TPC plants with IPP plants. Future research can be done to compare the efficiency among state-owned and IPP plants. In addition, similar approaches can be applied to analyze the productivity of other components of power systems (e.g., transmission and substation systems).


The authors would like to thank the generous assistance pro-vided by the Department of General Planning and Account-ing of Taiwan Power Company. Special thanks go to anony-mous reviewers for their constructive comments and invaluable suggestions.


[1] F. Y. Lo, “Develop a Technology-Productivity-Competitiveness-Oriented Operating Strategy for TPC,” Taipower Monthly Rep., vol. 474, pp. 33–36, 2002.

[2] Taiwan Power Company, 1999 Taiwan Power Company Annual Report. Taipei, Taiwan, R.O.C., 2000.

[3] A. Charnes, W. W. Cooper, and E. Rhodes, “Measuring the efficiency of decision making units,” Eur. J. Oper. Res., vol. 2, no. 6, pp. 429–444, 1978.

[4] R. W. Shephard, Theory of Cost and Production Functions. Princeton, NJ: Princeton Univ. Press, 1970.

[5] R. D. Banker, A. Charnes, and W. W. Cooper, “Some models for estimating technical and scale efficiencies in data envelopment analysis,” Manage.

Sci., vol. 30, no. 9, pp. 1078–1092, 1984.

[6] G. Debreu, “The coefficient of resource utilization,” Econometrica,

vol. 19, pp. 273–292, 1951.

[7] M. J. Farrell, “The measurement of productivity efficiency,” J. Royal Stat.

Soc., vol. 120, pp. 377–391, 1957.

[8] L. M. Seiford, “A bibliography for data envelopment analysis (1978– 1996),” Ann. Oper. Res., vol. 79, pp. 393–438, 1997.

[9] S. Gattoufi, M. Oral, and A. Reisman, “Data envelopment analysis lit-erature: A bibliography update (1951–2001),” Socio-Econ. Plan. Sci., vol. 38, pp. 159–229, 2004.

[10] A. Charnes, W. W. Cooper, A. Y. Lewin, and L. M. Seiford, Data

Envel-opment Analysis: Theory, Methodology and Application. Boston, MA: Kluwer, 1994.

[11] W. W. Cooper, L. M. Seiford, and K. Tone, Data Envelopment

Analy-sis: A Comprehensive Text with Models, Applications, and References.

Boston, MA: Kluwer, 2000.

[12] D. Caves, L. Christensen, and E. Diewert, “The economic theory of in-dex numbers and the measurement of input, output and productivity,”

Econometrics, vol. 50, no. 6, pp. 1393–1414, 1982.

[13] R. Fare, S. Grosskopf, M. Norris, and Z. Zhang, “Productivity growth, technical progress, and efficiency changes in industrialized countries,”

Amer. Econ. Rev., vol. 84, no. 1, pp. 66–83, 1994.

[14] R. Fare, S. Grosskopf, B. Lindgren, and P. Roos, “Productivity devel-opments in Swedish hospitals: A Malmquist output index approach,” in Data Envelopment Analysis: Theory, Methodology, and Applications, A. Charnes, W. W. Cooper, A. Y. Lewin, and L. M. Seiford, Eds. Boston, MA: Kluwer, 1995.

[15] P. L. Chang, S. N. Hwang, and W. Y. Cheng, “Using data envelopment analysis to measure the achievement and change of regional development in Taiwan,” J. Environ. Manage., vol. 43, pp. 49–66, 1995.

[16] T. Y. Chen and T. L. Yeh, “A measurement of bank efficiency, ownership and productivity changes in Taiwan,” Serv. Ind. J., vol. 20, no. 1, pp. 95– 109, 2000.

[17] R. Fare, S. Grosskopf, and J. Logan, “The relative efficiency of Illinois electric utilities,” Resour. Energy, vol. 5, pp. 349–367, 1983.

[18] A. Charnes, W. W. Cooper, D. Divine, T. W. Ruefli, and D. Thomas, “Comparisons of DEA and existing ratio and regression systems for ef-fecting efficiency evaluations of regulated electric cooperatives in Texas,”

Res. Gov. Nonprofit Account., vol. 5, pp. 187–210, 1989.

[19] P. A. Miliotis, “Data envelopment analysis applied to electricity dis-tribution districts,” J. Oper. Res. Soc., vol. 43, no. 5, pp. 549–555,


[20] F. Y. Lo, C. F. Chien, and J. T. Lin, “A DEA study to evaluate the relative efficiency and investigate the district reorganization of the Taiwan power company,” IEEE Trans. Power Syst., vol. 16, no. 1, pp. 170–178, Feb. 2001.

[21] C. F. Chien, F. Y. Lo, and J. T. Lin, “Using DEA to measure the relative efficiency of the service center and improve operation efficiency through reorganization,” IEEE Trans. Power Syst., vol. 18, no. 1, pp. 366–373, Feb. 2003.

[22] C. F. Chien, S. Chen, and Y. Lin, “Using Bayesian network for fault location on distribution feeder of electrical power delivery systems,” IEEE

Trans. Power Del., vol. 17, no. 3, pp. 785–793, Jul. 2002.

[23] J. Peng, C. F. Chien, and B. Tseng, “Rough set theory for data mining for fault diagnosis on distribution feeder,” Inst. Electr. Eng. Proc.-Gen.,

Trans. Distrib., vol. 151, no. 6, pp. 689–697, Nov. 2004.

[24] C. F. Chien, “Modifying the inconsistency of Bayesian network and a comparison study for fault location on electricity distribution feeder,” Int.

J. Oper. Res., vol. 1, no. 1/2, pp. 188–203, 2005.

[25] A. Pahwa, X. Feng, and D. Lubkeman, “Performance evaluation of electric distribution utilities based on data envelopment analysis,” IEEE Trans.


[26] A. D. Athanassopoulos, N. Lambroukos, and L. Seiford, “Data envelop-ment scenario analysis for setting targets to electricity generating plants,”

Eur. J. Oper. Res., vol. 115, pp. 413–428, 1999.

[27] T. Sueyoshi, “Tariff structure of Japanese electric power companies: An empirical analysis using DEA,” Eur. J. Oper. Res., vol. 118, pp. 350–374, 1999.

[28] S. U. Park and J. B. Lesourd, “The efficiency of conventional fuel power plants in South Korea: A comparison of parametric and non-parametric approaches,” Int. J. Prod. Econ., vol. 63, pp. 59–67, 2000.

[29] T. Sueyoshi and M. Goto, “Slack-adjusted DEA for time series analysis: Performance measurement of Japanese electric power generation industry in 1984–1993,” Eur. J. Oper. Res., vol. 133, pp. 232–259, 2001. [30] J. Raczka, “Explaining the performance of heat plants in Poland,” Energy

Econ., vol. 23, pp. 355–370, 2001.

[31] W. D. Cook and R. H. Green, “Evaluating power plant efficiency: A hierarchical model,” Comput. Oper. Res., vol. 32, pp. 813–823, 2005. [32] B. Golany, Y. Roll, and D. Rybak, “Measuring efficiency of power plants

in Israel by data envelopment analysis,” IEEE Trans. Eng. Manag., vol. 41, no. 3, pp. 291–301, Aug. 1994.

[33] J. D. Lee, J. B. Park, and T. Y. Kim, “Estimation of the shadow prices of pollutants with production/environment inefficiency taken into account: A nonparametric directional distance function approach,” J. Environ.

Manage., vol. 64, pp. 365–375, 2003.

[34] R. Fare, S. Grosskopf, D. W. Noh, and W. Weber, “Characteristics of a polluting technology: Theory and practice,” J. Econ., vol. 126, pp. 469– 492, 2005.

[35] B. Golany and Y. Roll, “An application procedure for DEA,” OMEGA, vol. 17, no. 3, pp. 237–250, 1989.

[36] Taiwan Power Company, Taiwan Power Company 2000 Annual Report. Taipei, Taiwan, R.O.C., 2001.

[37] Taiwan Power Company, The Implementation of Responsibility Center

System. Taipei, Taiwan, R.O.C., 1998.

[38] Taiwan Power Company, The Implementation of Profitability Center

Sys-tem. Taipei, Taiwan, R.O.C., 1998.

Chen-Fu Chien (M’03) was born in Taiwan, R.O.C.,

in 1966. He received the B.S. degree (with Phi Tao Phi Honor) in industrial engineering and electrical engineering from the National Tsing Hua Univer-sity (NTHU), Hsinchu, Taiwan, R.O.C., in 1990, the M.S. and Ph.D. degrees in industrial engineering from the University of Wisconsin – Madison, Madison, in 1994 and 1996, respectively.

During 2002–2003, he was a Fulbright Scholar in the Department of Industrial Engineering and Oper-ations Research, University of California, Berkeley, and during 2004–2005, he was a visiting Professor at Cambridge University (sponsored by the Royal Society). He is currently a Professor in the Department of Industrial Engineering and Engineering Management, NTHU. Since 2005, he has been on leave from NTHU to serve as a Deputy Director of Industrial Engi-neering Division, Taiwan Semiconductor Manufacturing Company. His current research interests include decision analysis, data mining, modeling and analysis for semiconductor manufacturing, and production strategy.

Dr. Chien is a member of the Institute of Industrial Engineers (IIE), the Insti-tute for Operations Research and the Management Sciences (INFORMS), and the Steering Committee of Industrial Engineering Division, National Science Council, Taiwan, R.O.C. He is also a Board Member of the Chinese Institute of Decision Sciences and Chinese Institute of Industrial Engineers (CIIE). He is the recipient of the Distinguished Young Industrial Engineer Award, the Dis-tinguished Young Faculty Research Award from NTHU, the Best Paper Award from CIIE, the Best Research Awards from the National Science Council, the Distinguished Industrial Collaboration Award from the Ministry of Education, the Best Engineering Paper Award from the Chinese Institute of Engineers, and the Tier 1 Principal Investigator (top 3%) of National Science Council (2005– 2008), Taiwan, R.O.C.

Wen-Chih Chen was born in Taiwan, R.O.C., in

1972. He received the B.S. degree in industrial en-gineering from the National Tsing Hua University, Hsinchu, Taiwan, R.O.C., in 1995, and the Ph.D. de-gree from the School of Industrial and Systems En-gineering, Georgia Institute of Technology, Atlanta, in 2003.

He is currently an Assistant Professor in the De-partment of Industrial Engineering and Management, National Chiao Tung University, Hsinchu. His current research interests include performance evaluation and decision-making analysis in logistics and supply chains.

Dr. Chen is a member of the Institute of Industrial Engineers (IIE) and the Institute for Operations Research and the Management Sciences (INFORMS).

Feng-Yu Lo was born in Taiwan, R.O.C., in 1968.

He received the Ph.D. dgeree in industrial engineer-ing and engineerengineer-ing management from the National Tsing Hua University, Hsinchu,Taiwan, R.O.C., in 2004.

He is a Manager in the Taiwan Power Company (TPC), Taipei, Taiwan, R.O.C. His current research interests include data envelopment analysis (DEA), decision analysis, and power industry.

Dr. Lo is a member of the Chinese Institute of Decision Sciences (CIDS) and the Chinese Institute of Industrial Engineers (CIIE).

Yi-Chiech Lin was born in Taiwan, R.O.C., in 1977. He received the M.S. degree

in industrial engineering and engineering management from the National Tsing Hua University, Hsinchu, Taiwan, R.O.C., in 2002. His current research interests include decision analysis, performance evaluation, and job assignment.

Mr. Lin is the recipient of the Best Thesis Award from the Chinese Institute of Decision Sciences, Taiwan, R.O.C.









相關主題 :