Energy-saving generation dispatch toward a sustainable electric power industry in China

Energy Policy 83 (2015) 14–25

Contents lists available at ScienceDirect

Energy Policy journal homepage: www.elsevier.com/locate/enpol

Energy-saving generation dispatch toward a sustainable electric power industry in China Haiwang Zhong a,n, Qing Xia a, Yuguo Chen b, Chongqing Kang a a b

Tsinghua University, Beijing, China Guangdong Power Grid Corporation, Guangzhou, China

H I G H L I G H T S








We propose a new type of Energy-Saving Generation Dispatch (ESGD) framework. Sequential coordination among different time-scale generation schedules. We propose SCUC and SCED models for the ESGD framework. Empirical analysis is conducted using the realistic data obtained from the Guangdong Power Grid. The existing financial interest distribution is maintained while energy savings are achieved.

art ic l e i nf o

a b s t r a c t

Article history: Received 20 November 2014 Received in revised form 10 March 2015 Accepted 15 March 2015

Energy shortages, climate change and environmental pollution are critical issues that the entire world is faced with currently. To tackle the challenge and realize sustainable development, the Chinese government launched the Energy-Saving Generation Dispatch (ESGD) in 2007. In the ESGD scheme, generating units are dispatched based on fuel consumption rates and pollutant emission intensities from low to high. However, annual generation quotas still widely exist. With the mandatory shutdown of smallcapacity and low-efficiency thermal generating units in 2006–2010, most of the currently running thermal generating units are large-capacity and highly efficient units. The additional improvement of the overall energy efficiency under this situation is a key problem for the Chinese electric power industry. To this end, a new type of ESGD framework is designed in this paper. Sequential coordination among yearly, monthly, day-ahead and real-time generation schedules is proposed. Based on the framework, the corresponding models are formulated. Empirical analysis is conducted using the realistic data obtained from the Guangdong Power Grid Corporation. Four generation dispatch modes are compared. The results indicate that the proposed ESGD mode can further reduce energy consumption and pollutant emissions. Hopefully, this paper can provide a valuable reference for policy making in the Chinese power sector. & 2015 Elsevier Ltd. All rights reserved.

Keywords: Energy-saving generation dispatch (ESGD) Sequential coordination Security constrained unit commitment (SCUC) Security constrained economic dispatch (SCED) CO2 emission reduction

1. Introduction 1.1. Background of energy-saving generation dispatch (ESGD) Energy shortages, climate change and environmental pollution are critical issues that the entire world is faced with currently. To tackle these challenges and realize sustainable development, many countries and regions have adopted policies to improve the energy efficiency and to increase the share of renewable energy (Lund et al., 2008).

n

Corresponding author. E-mail addresses: [email protected] (H. Zhong), [email protected] (Q. Xia), [email protected] (Y. Chen), [email protected] (C. Kang). http://dx.doi.org/10.1016/j.enpol.2015.03.016 0301-4215/& 2015 Elsevier Ltd. All rights reserved.

In China, coal plays a dominant role in the energy supply system, and this situation is likely to persist for a relatively long term. Since November 2013, the electricity generation of thermal power plants accounts for over 80% of the total electricity generation (China Electricity Council, 2013a). In 2012, approximately 3.64 billion tons of standard coal were consumed in China, and among that, 1.79 billion tons were utilized for electricity generation (China, 2012). In addition, among the CO2 emissions produced by fossil fuels, the power industry accounts for approximately 38% (Kang et al., 2012). The transportation sector is another large energy consumer. The energy consumption of transport sectors is about 20% of the total energy consumption in China (Wang et al., 2014). Therefore, from the perspective of sustainable development, effective actions must be taken to improve energy efficiency and to reduce CO2 emissions, such as integration of renewable

H. Zhong et al. / Energy Policy 83 (2015) 14–25

Nomenclature

N B L K T Ci(⋅) Pi, t Δt Ci , t , U Ci , t , D Pi, t Wt Dt α i, t

Pimin Pimax

Number of generating units Number of nodes Number of transmission lines Number of critical interfaces Number of periods Fuel consumption function of generating unit i Power output of generating unit i in period t Duration of each period Startup cost of generating unit i in period t Shutdown cost of generating unit i in period t Power output of unit i in period t Power output of renewable energy generation units System load demand in period t Binary variable that denotes the on/off states of unit i in period t . αi, t = 1 means the generating unit is on, while αi, t = 0 means the generating unit is off. Minimum stable generation of unit i Power rating of unit i

energy (Chen and Mei, 2015), carbon capture and storage (CCS) in the electricity sector (Chen et al., 2010), and electric vehicles (EVs) in the transportation sector (Ji et al., 2012; Hu et al., 2013; Yang et al., 2014; Ocran et al., 2005), respectively. This paper focuses on the electricity sector. In the past decades, to encourage investors to build power plants and to protect the interest of investors, the Chinese government implemented a generation dispatch mode termed “average generation and utilization hours” (Ding and Yang, 2013), which is also known as “percentage-based dispatch”. In the conventional generation dispatch mode, the annual utilization hours of similar types of generating units are nearly the same. Therefore, the annual generation quotas are nearly proportional to the unit capacity. However, this is obviously inefficient. For instance, the fuel consumption rate of an ultra-supercritical 1000-MW generating unit can be as low as 270 g of standard coal per kWh, while that of a 50-MW generating unit can be as high as 450 g of standard coal per kWh (Ding and Yang, 2013). Under these circumstances, the Chinese government launched the energy-saving generation dispatch (ESGD) in 2007 (NDRC, MEP, SERC, and NEA, 2007). In the Measures for Energy-Saving Generation Dispatching (Interim), the merit order table of the generating units in ESGD is shown in Table 1 (Gao and Li, 2010). Thermal generating units are dispatched based on fuel consumption rates from low to high. If the fuel consumption rates are

Table 1 The merit order table of the generating units in ESGD.

ΔPi, up ΔPi, dn Xion ,t−1 Xioff ,t−1 Tion Tioff STi SDi Rt PFk PFk Pl Pl Gl − i Di, t SSk Ei M Mi

15

Ramp-up limit of unit i Ramp-down limit of unit i ON time of generating unit i in period t − 1 OFF time of generating unit i in period t − 1 Minimum ON time of generating unit i Minimum OFF time of generating unit i Constant startup cost of generating unit i Constant shutdown cost of generating unit i System spinning reserve requirement in period t Positive transmission limits of interface k Negative transmission limits of interface k Positive transmission limits of line l Negative transmission limits of line l GSDF between node i and transmission line l Load at node i in period t Set containing the transmission lines in interface k Energy of generating unit i Days of this month, in this study M = 31 Days when the unit state is ON

Henan

Jiangsu

Sichuan Guizhou Guangdong

Fig. 1. The location of five ESGD pilot provinces in China.

identical, generating units will be further ranked by the pollutant emission intensities from low to high (Xu et al., 2009). The pilot work of ESGD was conducted in the provinces of Jiangsu, Henan, Sichuan, Guangdong and Guizhou in 2007 (Liu et al., 2010). The location of these provinces is shown in Fig. 1. Fuel consumption rate on-line monitoring systems are implemented in these pilot provinces, which can monitor the real-time fuel consumption of each thermal generating unit. The piloting experience showed that these pilot provinces, to some extent, reduced energy consumption and pollutant emissions (Ding and Yang, 2013). 1.2. New challenges emerge

Order Types of the generating units 1 2 3 4

5 6 7

Renewable generating units with little adjustment capability including wind, solar, oceanic energy and run-of-river hydro generating units Renewable generating units with adjusting capabilities including hydro generating units, biomass and geothermal energy generating units Nuclear generating units Combined cycle thermal generating units whose generation is decided by heating load and resource comprehensive exploitation generating units that specifically refer to waste power generating units Gas-fired generating units Other thermal generating units including a combined cycle thermal generating unit without heating load Oil-fired generating units

However, two challenges emerge with the currently implemented ESGD. First, because the implementation of ESGD results in the re-allocation of financial interest distribution, the reform encounters great resistance. The large-capacity and highly efficient generating units, renewable energy generating units, and hydropower generating units will obtain substantial benefit, while the small-capacity and low-efficiency generating units will be faced with heavy financial burden. In fact, the annual generation quotas still widely exist. Additionally, the current generation dispatch mode focuses on the day-ahead generation schedules. The generation company is reluctant to shut down the generating unit

16

H. Zhong et al. / Energy Policy 83 (2015) 14–25

Fig. 2. Average fuel consumption rate curves of thermal units.

because once the generating unit is shutdown, the startup time is uncertain. In this mode, the generation company is faced with financial risks, and it is difficult for power plants to manage the production plan in a relatively long term. Second, in the 11th Five-Year Plan1 period (2006–2010), China has shutdown small-capacity and low-efficiency thermal generating units with a total capacity of 76.83 GW. In 2012, the national average of coal consumption for electricity generation decreased to approximately 325 g/kWh, catching up with the advanced world level (China Electricity Council, 2013b). However, with the mandatory shutdown of small-capacity and low-efficiency thermal generating units, most of the currently running thermal generating units are large-capacity and highly efficient thermal generating units. According to the on-line monitored data from the ESGD pilot provinces, the average fuel consumption rate of the coal-fired thermal generating units are descending, as shown in Fig. 2. As one can observe, the higher the loading rate is, the higher the efficiency is, i.e., the lower the average fuel consumption rate is. If these generating units are operated in the percentage-based generation dispatch mode as before, the loading rates for on-line generating units would be relatively low. In that case, the overall energy efficiency is low. The further improvement of the overall energy efficiency under this situation is a key problem for the electric power industry in China. In response to these challenges, a new ESGD mode is designed in this paper. In the proposed ESGD mode, the yearly, monthly, day-ahead and real-time power generation schedules can be effectively coordinated in the time dimension. The power resources allocation can be optimized in a longer time horizon. Most importantly, the potential benefit of energy savings and emission reduction for the generation schedules in power systems can be further obtained. The rest of this paper is organized as follows: Section 2 describes the methods of the proposed ESGD mode in which the sequential coordination and feedback mechanism will be discussed. In addition, the mathematical models are formulated, and the corresponding solution methodology is proposed. In Section 3, the empirical analysis on ESGD in the Guangdong Power Grid including the basic data, indices and results are presented. Section 4 presents the discussion. Section 5 provides the conclusion, and several policy suggestions are proposed. 1 The Five-Year Plans are a series of initiatives that guide the social and economic development of China for the next five years.

2. Methods As mentioned in Section 1, the annual generation quotas still widely exist in China, and the current generation dispatch mode focuses on the day-ahead generation schedules. In this mode, it is difficult for power plants to manage their production plans in a relatively long term. To resolve this drawback, a new ESGD mode is designed in this section. In the new ESGD mode, to optimize the allocation of power resources in a longer time horizon, yearly, monthly, daily and real-time generation schedules are sequentially coordinated. On one hand, the upstream generation schedules (e.g., yearly and monthly generation schedules) provide the optimization boundary, adequate decision space and adjustment capability for downstream generation scheduling (e.g., day-ahead and real-time dispatch). From the upstream generation schedule to the downstream generation schedule, information such as load forecast and grid topology becomes more accurate, and the time resolution becomes smaller, e.g., from a month to 15 min or 5 min. On the other hand, according to the actual operation, the deviation between the actual implementation and schedules is sent back to the upstream generation scheduling modules as feedback information. The downstream generation schedule discovers the deviation and critical constraints that should be taken into account in the upstream generation scheduling models. In this way, the parameters of the scheduling models can be adaptively and continuously improved. The closed-loop and self-adaptive ESGD mode is depicted in Fig. 3. The key feature of the proposed ESGD mode is the sequential coordination and feedback mechanism. In the sequel, the details of the proposed ESGD mode will be explained. 2.1. Sequential coordination process of generation scheduling 2.1.1. Yearly generation schedule The objective of the yearly generation schedule is: at the end of each year, according to the next year's electricity production plan, monthly load forecast, electricity interchange with other regions (through tie-lines), and generation maintenance schedules, the annual electricity quota is allocated into every month of the next year. The yearly generation scheduling is conducted in a rolling manner. The allocation of the monthly electricity quantity is adjusted after each month's operation based on the rate of fulfillment of the yearly generation schedule and generation maintenance schedule in the remaining months.

H. Zhong et al. / Energy Policy 83 (2015) 14–25

Adjust Generation Maintenance

Yearly Generation Schedule

Generation Maintenance Electricity Production Allocation

17

Adjust Unit Power Output

Monthly Generation Schedule

Unit Commitment

Day-ahead Generation Schedule

Unit Power Output

Real-time Generation Dispatch

Adjust Unit Commitment

Fig. 3. Sequential coordination of yearly, monthly, day-ahead and real-time generation schedules.

2.1.2. Monthly generation schedule The objective of the monthly generation schedule is: at the end of each month, according to the next month's load forecast and generation maintenance schedules, the monthly unit commitment is optimized. The objective of unit commitment is to decide the on/off states of generating units (Wood and Wollenberg, 1996). The constraints considered include: the load balance, system reserves, capacity of generating units, minimum on/off time limit (Bard, 1988), and critical transmission interface2 limit constraints. 2.1.3. Day-ahead generation schedule The objective of the day-ahead generation schedule is: in dayahead, given the pre-determined monthly unit commitment result, the power output profile of each generating unit during the next day is optimized to balance the electricity demand. The optimization is conducted according to the latest information including the grid topology, day-ahead system load forecast and nodal load forecast. The constraints include: the load balance, capacity of generating units, ramping limits of generating units, transmission line and transformer capacity, and interface transmission constraints. 2.1.4. Real-time generation dispatch One of the salient features of power systems is the instantaneous balance of power generation and load demand. Otherwise, the frequency of electricity will deviate from the nominal value (e.g., 50 Hz in China and 60 Hz in the U.S.). According to the deviation between the near real-time forecast of operational conditions and the day-ahead forecast, the power output of units is adjusted to maintain an instantaneous load balance and to ensure the security of the power systems. From yearly to monthly, from monthly to day-ahead, and from day-ahead to real-time, the forecast information becomes more accurate, and the uncertainties are gradually reduced. Through optimal decision making based on more accurate information, the proposed ESGD effectively and sequentially coordinates these generation schedules on different time scales. Therefore, more energy savings and emission reduction benefits can be captured, while the security of the power grids remains assured. 2.2. Feedback mechanism of generation scheduling In the long-term and mid-term generation schedules (e.g., yearly and monthly), because the decision maker is faced with many uncertainties, it is difficult to predict all of the critical constraints that may become binding in actual operations. Hence, a feedback mechanism is required to send the deviation information back to the upstream scheduling modules. In this way, the parameters of the generation scheduling models can be modified in a timely manner and can be continuously updated. 2.2.1. Feedback from monthly operations to yearly generation scheduling The updated information including the load forecast error, the temporary change or cancellation of generation maintenance 2 A transmission interface is a transmission line group or a path that consists of several transmission lines.

schedules should be sent back to the yearly generation scheduling module as feedback to properly adjust the allocation of the electricity quantity in the remaining months. 2.2.2. Feedback from daily operations to monthly generation scheduling By analyzing the actual operation data, the transmission lines and interfaces that are frequently congested can be sent back to the monthly generation scheduling module. In this way, the monthly unit commitment can be more implementable in day-ahead and real-time operations, and the security of the power system is improved. 2.2.3. Feedback from real-time operations to day-ahead generation scheduling In the real-time operation of power systems, the system condition may deviate from the forecast because of the load forecast error and forced outage of generating units. The deviation between the actual operating point and the day-ahead forecast should be sent back to the day-ahead generation scheduling module. In this way, the deviation between the day-ahead schedule and the actual operations can be minimized. 2.3. The advantage of the proposed ESGD mode Compared with the currently implemented ESGD mode that merely emphasizes day-ahead generation scheduling, the proposed ESGD mode can realize: 1. Smooth electricity production. Because the generation schedules on different time scales can be coordinated, power resource allocation can be achieved in a longer time period. For instance, the rate of fulfillment of the yearly electricity generation quantity is monitored through the year. At the end of each month, the rate of fulfillment of the yearly electricity generation quantity is checked for each unit. If the rate of fulfillment is lagging, the unit will be allocated more electricity generation quantity in the following months, and vise versa. Similarly, in each day, the rate of fulfillment of the monthly electricity generation quantity is checked for each unit. If the rate of fulfillment is lagging, the unit will be allocated more electricity generation quantity in the following days, and vise versa. In this way, for power producers, a reasonable production plan can be made in advance. The annual generation quotas of generating units can be completed in a smoother way. 2. Energy savings and emission reduction of power system operations while maintaining the existing financial interest distribution among stakeholders. In the currently implemented ESGD mode, the day-ahead generation schedule is emphasized, and generation companies are reluctant to shut down the generating unit because once the generating unit is shutdown, the generation company does not know when the generating unit can be started up again. However, in the proposed ESGD mode, unit commitment is optimized in a monthly horizon. The generation company can be informed of the startup and shutdown schedules a month ahead. The shutdown of some of the generating units can increase the loading rates of on-line generating units. Therefore, the efficiency of on-line generating units is improved, and the overall efficiency is also improved.

18

H. Zhong et al. / Energy Policy 83 (2015) 14–25

To realize the above ESGD mode, mathematical models need to be established. In the sequel, the monthly and day-ahead generation scheduling models are focused on. The monthly generation scheduling model is a security constrained unit commitment (SCUC) model (Shahidehpour et al., 2002, p. 275; Fu and Shahidehpour, 2007), while the day-ahead generation scheduling model is a security constrained economic dispatch (SCED) model (Vargas et al., 1993).

period t − 1. Tion is the minimum ON time of generating unit i . Tioff is the minimum OFF time of generating unit i . 4. Ramping constraints Let βi, t = αi, t − αi, t − 1. Then, the ramping constraints can be formulated as follows:

Pi, t − Pi, t − 1 ≤ αi, t − 1ΔPi, up + βi, t Pimin + Pimax(1 − αi, t), t

Monthly generation scheduling is a key procedure in the proposed ESGD mode. The objective of monthly generation scheduling is to minimize the system operating cost while satisfying all types of constraints including the system-wide and unit-wise constraints (Fu and Shahidehpour, 2007). 2.4.1. Objective function The objective function of monthly generation scheduling is to minimize the operating costs (mainly fuel consumption), startup costs and shutdown costs (Carrion and Arroyo, 2006), which can be expressed as follows:

min F =

N

⎡

T

(6)

= 2, …, T

2.4. Monthly generation scheduling

⎤

∑ n= 1 ⎢⎣ ∑t = 1 (Ci(Pi, t)Δt + Ci, t, U + Ci, t, D)⎥⎦

(1)

In (1), N is the number of generating units and T is the number of periods. Ci(⋅) is the fuel consumption function of generating unit i . Pi, t is the power output of generating unit i in period t . Δt is the duration of each period. Ci, t , U is the startup cost of generating unit i in period t . Ci, t , D is the shutdown cost of generating unit i in period t. 2.4.2. Constraints The following constraints are considered (Arroyo and Conejo, 2000; Arroyoand Conejo, 2004): 1. Power balance

Pi, t − 1 − Pi, t ≤ αi, t ΔPi, dn + βi, t Pimin + Pimax(1 − αi, t − 1), t (7)

= 2, …, T

The ramping constraints (6) prevent a startup unit from producing more than Pimin MW in its first period on-line. The ramping constraints (7) prevent a shutdown unit from producing more than Pimin MW in its last period on-line. 5. Startup cost Similarly, the startup cost merely relies on the on/off state variable of generating units, αi, t . STi is the constant startup cost of generating unit i .

Ci, t, U ≥ STi(αi, t − αi, t − 1), t = 2, …, T

(8)

Ci, t, U ≥ 0, t = 1, …, T

(9)

6. Shutdown cost Similarly, the shutdown cost merely relies on the on/off state variable of generating units, αi, t .SDi is the constant shutdown cost of generating unit i .

Ci, t, D ≥ SDi (αi, t − 1 − αi, t), t = 2, …, T

(10)

Ci, t, D ≥ 0, t = 1, …, T

(11)

7. System Spinning Reserve Constraints N

∑i = 1 Pi, t + Wt = Dt , t = 1, 2, …, T

(2)

The electricity supply has to balance the electricity demand. Pi, t is the power output of unit i in period t . Wt is the output of renewable energy generation units. Dt is the system load demand in period t . Transmission loss is not considered in this study. 2. Generation limits

αi, tPimin ≤ Pi, t ≤ αi, tPimax, i = 1, 2, …, N; t = 1, 2, …, T

(

Tion

(3)

N

B

∑ ∑i = 1 Gl − iPi, t − ∑ ∑i = 1 Gl − iDi, t

l ∈ SSk

l ∈ SSk

¯ ,k ≤ PF k

(αi, t − 1 − αi, t) ≥ 0, t = 2, …, T

(4)

(Xioff,t − 1 − Tioff )

(αi, t − 1 − αi, t) ≥ 0, t = 2, …, T

(5)

In (4) and (5), Xion , t − 1 is the ON time of generating unit i in period t − 1. Xioff , t − 1 is the OFF time of generating unit i in

8. Interface power flow constraints Because of the low-frequency oscillation and voltage stability caused by long distance transmission lines, the transmission limits of the interfaces need to be considered. DC power flow is adopted in the monthly generation scheduling model. To promote the modeling efficiency (Biskas et al., 2005), the power flow constraints are expressed by the generation shift distribution factor (GSDF) (Ng, 1981).

PF̲ k ≤

)

−

(12)

In (12), Rt is the system spinning reserve requirement in period t .

In (3), αi, t is a binary variable that denotes the on/off states of unit i in period t . αi, t = 1 means the generating unit is on, while αi, t = 0 means the generating unit is off. Pimin and Pimax are the minimum stable generation and power rating of unit i , respectively. 3. Minimum up/down time limits

Xion ,t−1

N

∑i − 1 αi, tPimax ≥ Dt + Rt , t = 1, 2, …, T

= 1, 2, …, K , t = 1, …, T

(13)

In (13), K is the number of critical interfaces that are considered in the monthly generation scheduling procedure. These interfaces are potentially congested in the

H. Zhong et al. / Energy Policy 83 (2015) 14–25

19

operations. These interfaces are updated continuously via the feedback mechanism designed in Section 2.2. Pk and Pk are the negative and positive transmission limits of interface k . N is the number of generators. B is the number of nodes. Gl − i is the GSDF between node i and transmission line l . Di, t is the load at node i in period t . SSk is the set that contains the transmission lines in interface k . 9. Unit energy constraints To maintain the current financial interest distribution among generation companies, the unit energy constraints are incorporated in the monthly generation scheduling model. T

∑t = 1 Pi, t = Ei, i = 1, 2, …, N

(14)

In (14), Ei is the energy contract of generating unit i , which reflects the “average generation and utilization hours” principle. Equations (1)–(14) formulate the monthly generation scheduling model. It is a mixed-integer non-linear programming (MINLP) problem. It is worth noting that other constraints such as emission constraints can be included if desired.

Fig. 4. Piecewise linearization of the total fuel consumption curve.

In (20), L is the number of transmission lines. 5. The transmission interface power flow constraint and unit energy constraints are similar to those in the monthly generation scheduling model.

2.5. Day-ahead generation scheduling 2.5.1. Objective function Given the on/off states of the generating units resulted from the monthly generation schedules, the objective function of day-ahead generation scheduling is to minimize the operating cost (mainly fuel consumption).

min F =

∑ n= 1 ∑t = 1 ⎡⎣Ci(Pi, t)Δt ⎤⎦ N

T

(15)

2.5.2. Constraints The following constraints are considered: 1. Power balance constraints are similar to those in the monthly generation scheduling model. 2. Generation limits For units that are determined to be on, the constraints are as follows:

Pimin ≤ Pi, t ≤ Pimax, i = 1, 2, …, N; t = 1, 2, …, T

(16)

For units that are determined to be off, the constraints are as follows:

Pi, t = 0, i = 1, 2, …, N; t = 1, 2, …, T

(17)

3. Ramping constraints

Pi, t − Pi, t − 1 ≤ ΔPi, up, i = 1, 2, …, N; t = 1, 2, …, T

(18)

Pi, t − 1 − Pi, t ≤ ΔPi, dn, i = 1, 2, …, N; t = 1, 2, …, T

(19)

4. Line power flow constraints DC power flow is adopted in the day-ahead generation scheduling model.

P̲ l ≤

The above objective function and constraints formulate the day-ahead generation scheduling model. Because the on/off states of the generating units are determined in the monthly generation scheduling, there is no binary variable in the day-ahead generation scheduling model. Therefore, it is a non-linear programming (NLP) problem.

N

B

∑i = 1 Gl − i Pi, t − ∑i = 1 Gl − i Di, t

≤ P¯l, l = 1, …, L; t = 1, 2, …, T

(20)

2.6. Fuel consumption curve As mentioned previously, the fuel consumption rate of coalfired thermal generating units are descending. The objective of the ESGD model is to minimize the total fuel consumption of thermal generating units. Because it is difficult to build a mathematical model based on the descending average fuel consumption rate curves, the total fuel consumption curves in Fig. 4 are adopted in the formulation. The point (Hi, Pi) at the curve refers to the total fuel consumption per hour Hi at the output Pi . The horizontal axis refers to the output of units (MW), and the vertical axis refers to the total fuel consumption per hour (ton/h). The total fuel consumption curve is ascending. It can be fitted using a quadratic function.

H = aP 2 + bP + c

(21)

In practice, the coefficients of the quadratic term are typically positive. Therefore, the total fuel consumption curves are convex, and the ESGD model is a convex MINLP problem. In this paper, a piecewise linearization method (Fletcher, 1987, p. 357; Zimmerman, et al., 2011) is utilized to approximate the quadratic total fuel consumption curve. A series of linear functions are introduced to convert the model to a mixed integer linear programming (MILP) problem. In Fig. 4, mi is the slope of the ith piecewise linear segment, while ni is the intercept of the ith piecewise linear segment. The prevailing approach to solve the SCUC problem is Lagrangian Relaxation (LR) and mixed integer programming (MIP). The LR approach decomposes the SCUC problem into a series of single unit sub-problems (Wang et al., 1995). However, it is difficult to handle the coupling constraints among units. Given the sophisticated constraints in the SCUC model, MIP is selected as the solution approach. The prevailing approach to solve the SCED problem is quadratic programming (QP) and linear programming (LP) (Xie and Ilic, 2010). There are several off-the-shelf MIP/QP/LP

20

H. Zhong et al. / Energy Policy 83 (2015) 14–25

Table 2 Basic data of the Guangdong Power Grid. Characteristic quantity

Value (MW)

Generating unit type

Number

Capacity (MW)

Network topology

Number

Maximum load Minimal load Average load Maximum exchange of tie-lines Maximum load of local unitsa Maximum total load

38,777 25,527 32,716 17,000 4800 60,577

Coal Hydropower Gas Nuclear Pumped storage Oil

117 16 25 5 8 36

37,592 875 6016 4050 1800 3000

Node Line Transformer Interface

480 878 85 102

a

Local units refer to those generating units that are not dispatched by provincial power dispatch center.

solvers, such as The IBM CPLEX Optimizer (2014) and Gurobi Optimizer (2014). 2.7. Data acquisition The actual operational data obtained from the Guangdong Power Grid Corporation are utilized to verify the energy-saving effect of the proposed ESGD mode and model and to analyze the energy-saving potential of the Guangdong Power Grid. The characteristic indices of the system load, the number and capacity of different types of units, and the information for the network topology in the Guangdong Power Grid is summarized in Table 2. The system load profile (tie-line exchange excluded) of the Guangdong Power Grid is shown in Fig. 5. 2.7.1. Generation mix Generating units in the Guangdong Power Grid mainly include coal-fired thermal generating units, gas-fired thermal generating units and nuclear generating units; the number of hydropower generating units is small. 2.7.2. Fuel consumption characteristics of thermal generating units The total fuel consumption rate curves of the thermal generating units are shown in Fig. 6. They are ascending convex curves. 2.7.3. Startup and shutdown cost of thermal generating units Startup costs vary from unit to unit. In this study, for simplicity, startup costs of generating units with the same capacity are assumed to be the same, and startup costs of generating units with different capacities are assumed to be proportional to the capacity. According to empirical data, a 600-MW unit has a startup cost of approximately 600,000 RMB, while a 300-MW unit has a startup cost of approximately 300,000 RMB. Because the objective function is fuel consumption, the startup costs are converted to the unit of fuel consumption. In 2012, the price of raw coal is approximately 700–800 RMB/ton (Bohai-rim Steam-Coal Price Index, 2012), and the conversion coefficient between raw coal and standard coal is 0.7143 (1 kg raw coal ¼ 0.7143 kg standard coal) (Chinese National Standard, 2008), so the

price of standard coal is 980–1120 RMB/ton. In this paper, it is considered to be 1000 RMB/ton. For simplicity of analysis, the shutdown costs of thermal units are not considered in this study.

3. Results The simulation platform is developed using the C language with Microsoft Visual Studio. CPLEX is adopted to solve the MILP and LP models. 3.1. Scenarios Four scenarios are set up to analyze the energy-saving benefits in different generation dispatch modes. (1) Scenario 1: Percentage-based generation dispatch This is the conventional generation dispatch mode, i.e., average generation and utilization hours. The objective of generation dispatch is to maintain equity among units. In Scenario 1, according to the percentage-based generation dispatch principle, monthly generation schedules are rearranged based on the actual system load information, aiming at the “average monthly generation and utilization hours” among different units. (2) Scenario 2: Actual generation dispatch result This scenario is to replicate the actual operations of the power system. The indices are calculated based on the historical data obtained from the Guangdong Power Grid Corporation, including the unit commitment states, the power output of generating units, and the load demand profiles. (3) Scenario 3: Unit commitment optimization with energy constraints In this scenario, the electricity quantity of each unit is first calculated based on the historical data. To maintain the current financial interest distribution among generating units, these actual electricity quantities are imposed as constraints in the monthly generation scheduling model. This scenario is used to study the amount of energy savings benefit that can be

Fig. 5. System load profile of Guangdong Power Grid on a typical day.

H. Zhong et al. / Energy Policy 83 (2015) 14–25

21

Fig. 6. Total energy consumption curves for generating units with different capacities.

achieved without changing the current financial interest distribution, which reflects a gradual reform strategy. This scenario represents the proposed ESGD mode. (4) Scenario 4: Unit commitment optimization without energy constraints In this scenario, the electricity quantity constraints of units are not considered. This scenario represents the radical transition from the conventional generation dispatch mode to the ideal ESGD mode, which is used to demonstrate the maximum energy-saving potential. The quadratic fuel consumption curves are considered in the objective function. 3.2. Indices for empirical analysis To compare different generation dispatch modes, the following indices are analyzed in this study.


Total quantity of saved standard coal










Based on the conventional percentage-based generation dispatch scenario, the total quantity of saved standard coal of each scenario can be calculated. Average saved standard coal Based on the total quantity of saved standard coal and the total electricity quantity, the average saved standard coal of each scenario can be calculated. Percentage of saved standard coal Based on the conventional percentage-based generation dispatch scenario, the percentage of saved standard coal of each scenario can be calculated. Daily average loading rate of thermal units The loading rate of a thermal unit i refers to the ratio between power output and capacity.

ηi =

Pi Pimax

(22)


Total fuel consumption of all thermal generating units




Based on the output profile and the actual fuel consumption rate curve of thermal units, total fuel consumption of thermal units can be calculated. Average fuel consumption of thermal generating units According to the total fuel consumption and total electricity quantity of thermal units, the average fuel consumption of all thermal units can be calculated.

The daily average loading rate of a thermal unit i is defined as: T

η¯i, day =

∑t = 1 Pi, t T × Pimax

(23)

where T refers to the total periods of one day. In this study, T =24 ; Pi, t refers to the output of unit i in period t ; Pimax refers to

22

H. Zhong et al. / Energy Policy 83 (2015) 14–25

Table 3 Energy consumption statistics of each scenario. Total fuel consumption Scenario (thousand tons standard coal)

Average fuel consumption (g/kWh)

Saved standard coal (thousand tons)

Average saved standard coal (g/ kWh)

Percentage of saved standard coal (%)

1 2 3 4

311.09 306.51 304.14 294.25

–

– 4.58 6.95 16.84

– 1.47 2.23 5.41

5452 5372 5330 5157

80 122 295

4. Discussion

the capacity of unit i . If the unit state is OFF on the day, η¯i, day = 0.

The monthly average utilization rate of a thermal unit i is defined as: T ×M

∑t = 1 Pi, t M × T × Pimax

(24)

This index reflects the ratio between the monthly electricity quantity and the maximum generating capacity of the unit. M refers to the days of this month, which in this study is M = 31. If the unit has been shut down during the whole month, η¯i, month = 0.


Monthly average operational loading rate of thermal unit The monthly average operational loading rate of a thermal unit i is defined as: T ×M

ρi, month =

∑t = 1 i Pi, t Mi × T × Pimax

(25)

where Mi refers to the days when the unit state is ON. Therefore, this index only reflects the average loading rate during the operational periods, excluding the shutdown periods. If the unit has been shut down during the whole month, ρi, month = 0.


Total electricity purchasing cost of thermal units




In this section, the energy-saving benefits of each scenario are compared and analyzed with respect to the percentage-based generation dispatch mode (Scenario 1). 4.1. Comparisons of fuel consumption


Monthly average utilization rate of thermal units

η¯i, month =

purchasing cost of the thermal units can be calculated. The results of each scenario are shown in Table 3. It is worth noting that most of the generating units in the Guangdong Power Grid are large-capacity high-efficiency units; therefore, the average fuel consumption rate is lower than the national average level (325 g/kWh) in China.

According to the electricity tariff and the electricity quantity of each thermal unit, the total electricity purchasing cost of the thermal units will be summed to reflect the economy of the generation schedules. Average electricity purchasing cost of thermal units

According to the total electricity purchasing cost and the total electricity quantity of the thermal units, the average electricity

In Scenario 1, the conventional generation dispatch mode aims at maintaining equity among generating units, therefore the utilization hours of the generating units are nearly the same. In this scenario, large-capacity high-efficiency generating units are underutilized. In this scenario, most large-capacity units operate during the whole month, while the operational loading rate is relatively low. As a result, the fuel consumption is the highest among all of the scenarios. In Scenario 2, the actual generation dispatch schedule has optimized the unit commitment and output and has saved energy to some extent. However, the current generation dispatch mode merely emphasizes the day-ahead generation schedules. Monthly and day-ahead generation schedules are not effectively coordinated. Therefore, the energy-saving potential can be further promoted. In Scenario 3, with the premise of ensuring the monthly electricity quantity of each unit through the cyclic operation among thermal units, energy savings have been realized by promoting the operational loading rate. The salient feature of this scenario is that it does not change the current financial interest distribution. Compared with Scenario 2, the fuel consumption per kWh is reduced by 2.37 g of standard coal. This part of the energy-saving benefit is brought on by the proposed ESGD mode. In Scenario 4, the outputs of large-capacity units are considerably close to their capacities. Generating units with a capacity of less than 300 MW have less utilization hours and may be shut down for a period of time in the studied month. However, in this scenario, the electricity quantities of the thermal units are only determined by the generation merit order table, which is a radical transition path towards ESGD. In this scenario, the theoretical maximum energy-saving potential is provided. In terms of energy-saving benefits, in Scenario 3, the energysaving benefits are mainly achieved by the cyclic operation of generating units with the premise of ensuring the monthly

Fig. 7. Monthly average utilization rate curve in Scenario 1.

H. Zhong et al. / Energy Policy 83 (2015) 14–25

23

Fig. 8. Monthly average utilization rate curve in Scenario 4. Table 4 Electricity purchasing costs of each scenario. Scenario Total electricity pur- Average electricity purchasing cost chasing cost (bil. (RMB/MWh) RMB)

Average energy consumption (g/kWh)

1 2 3 4

311.09 306.51 304.14 294.25

7.725 7.733 7.733 7.680

440.80 441.25 441.25 438.23

electricity quantity of each generating unit, promoting their operational loading rates. In Scenario 4, the energy-saving benefits are realized mainly by the substantial adjustment of the utilization hours among generating units. Although scenario 4 acquires more substantial energy-saving benefits because the financial interest balance is broken, it will encounter more resistance from the stakeholders in actual implementation.

descends according to the merit order table of the generating units. Additionally, the small-capacity units at the bottom of the merit order table generate no electricity, and the unit state is OFF during the whole month. Overall, in Scenario 1, the difference in the utilization rate between large-capacity and small-capacity generating units is the smallest. Therefore, the fuel consumption in this scenario is the highest. In Scenario 4, the difference in the utilization rate between large-capacity and small-capacity generating units is the largest. Therefore, the fuel consumption is the lowest. Hence, this conclusion can be drawn: the larger the difference in utilization rates between large-capacity and small-capacity generating units is, the more energy-saving benefits can be obtained. The result of Scenario 3 illustrates that even when the energy constraints (utilization hours) of generating units are restricted, energy-saving benefits can still be obtained via a coordinated cyclic operation of generating units. The underlying reason is that by raising individual loading rates, the efficiency of thermal units can be improved.

4.2. Comparisons of utilization hours 4.3. Comparisons of electricity purchasing costs To analyze the adjustment of the utilization hours in Scenario 4, the monthly average utilization rate curves in Scenario 1 (percentage-based generation dispatch) and Scenario 4 are shown in Figs. 7 and 8, respectively. The horizontal axis refers to the merit order table of generating units. The capacity of units descends from left to right. The vertical axis refers to the monthly average utilization rate of the corresponding units. In Scenario 1, the utilization rates of all thermal units are nearly the same, and the financial interest distribution of different generating units is the most equitable. However, the fuel consumption in this scenario is the highest. In Scenario 4, the energy constraints are not taken into consideration. The monthly average utilization rate of the unit

The electricity purchasing costs of each scenario are shown in Table 4. According to the actual situations of the Guangdong Power Grid, the electricity tariffs of small-capacity units are generally higher than those of the large-capacity units. The electricity purchasing cost in Scenario 4 is the lowest. In Scenarios 2 and 3, the energy constraints of generating units are the same; therefore, the electricity purchasing costs are the same. 4.4. Cyclic operation of thermal units The cyclic operation benefits refer to energy savings through the cyclic shutdown of thermal units which results in an increase

Fig. 9. Actual daily average loading rate curves of Mawan #1, #2, and #3 units.

24

H. Zhong et al. / Energy Policy 83 (2015) 14–25

Fig. 10. Optimized daily average loading rate curves of Mawan #1, #2, and #3 units.

of their operational loading rates. For instance, the actual daily average loading rate curves of Mawan #1, #2, and #3 units are shown in Fig. 9, and the optimized daily average loading rate curves of Mawan #1, #2, and #3 units in Scenario 3 are shown in Fig. 10. As one can observe, Mawan #1, #2, and #3 units shutdown in period 17–21, 7–11, 10–17, respectively. By cyclic shutdown, the average operation loading rates of the units are increased. It is worth noting that there are some limitation for this study: 1. Only the electricity sector is investigated in this paper. In fact, the transportation sector is also an important aspect for energy savings. 2. The renewable energy is considered in a simple way in this paper. In fact, with the increase of the renewable energy penetration, the power system operations will face great challenges. More complicated models and methods are needed to consider the stochastic and intermittent characteristics for renewable energy. These can be investigated in our future work.

5. Conclusions and policy implications This paper presents a new ESGD mode. In the proposed ESGD mode, the yearly, monthly, day-ahead and real-time generation schedules are sequentially coordinated in a rolling manner. The sequential coordination process and the feedback mechanism are designed, respectively. The monthly and day-ahead generation scheduling models and efficient algorithms are proposed. Based on the models and algorithms, the simulation platform is developed. Based on the realistic operational data obtained from Guangdong Power Grid Corporation, four scenarios are set up and analyzed on the simulation platform. In terms of the fuel consumption, utilization rate, operational loading rate, cyclic operation and electricity purchasing cost of the units, four scenarios are well analyzed and compared to illustrate the superiority of the ESGD mode and model proposed in this paper. The empirical analysis shows that the proposed ESGD mode can provide an additional energysaving benefit, given that the annual electricity quota of the generating units is unchanged. In other words, the proposed ESGD mode can further improve the overall energy efficiency for the electric power industry, given that the financial interest distribution is unchanged. Based on the theoretical analysis and empirical results, this study has several important policy implications as follows: First, the currently implemented ESGD puts emphasis on dayahead generation scheduling. It can be changed into the proposed

ESGD mode. The power resource allocation can be optimized in a longer time horizon. The salient advantage of the proposed ESGD is that gradual reform can be realized because the financial interest distribution is maintained. Therefore, the policy is more acceptable for stakeholders. From the societal perspective, more benefits from energy savings and emission reduction can be grasped. Second, it would be sensible to conduct some pilot projects before developing a wide application of the proposed ESGD in China. Although the proposed ESGD already shows promising results using the realistic data obtained from Guangdong Power Grid Corporation on the developed simulation platform, more experience can be accumulated through pilot projects. Some hidden problems may be found and fixed during the pilot practice. Third, an education program for power plants should be conducted. One of the barriers that prevent the implementation of ESGD is that generation companies are reluctant to shutdown generating units. The reason is that once the generating unit is shutdown, the startup time is uncertain. The government can explain the benefit of the monthly generation scheduling and cycling strategy to generation companies. Therefore, it would be easier to implement the ESGD policy. This study may provide a reference for policy making in the Chinese government.

Acknowledgments This research was supported by the National Natural Science Foundation of China (Nos. 51325702, 51407100) and the China Postdoctoral Science Foundation (No. 2014M560969)

References Lund, H., Kempton, W., 2008. Integration of renewable energy into the transport and electricity sectors through V2G. Energy Policy 36 (9), 3578–3587. China Electricity Council, 2013a. China's Electric Power Industry Profiles, January– November 2013.〈http://www.cec.org.cn/guihuayutongji/gongxufenxi/dia nliyunxingjiankuang/2013-12-17/113992.html〉, (accessed 13.09.14.). China Statistical Yearbook, 2012. China Statistics Press, 2012, Beijing. Kang, C., Zhou, T., Chen, Q., 2012. Carbon emission flow in networks. Sci. Rep. 2, 479 〈http://dx.doi.org/10.1038/srep00479〉. Wang, Y., Li, K., Xu, X., et al., 2014. Transport energy consumption and saving in China. Renew. Sustain. Energy Rev. 29, 641–655. Chen, L., Mei, S., 2015. An Integrated Control and Protection System for Photovoltaic Microgrids, CSEE Journal of Power and Energy Systems 1 (1), 36–42. Chen, Q., Kang, C., Xia, Q., et al., 2010. Power Generation expansion planning model towards low-carbon economy and its application in China. IEEE Trans. Power Syst. 2, 1117–1125. Ji, S., Cherry, C.R.,J., Bechle, M., Wu, Y., Marshall, J.D., 2012. Electric vehicles in China: emissions and health impacts. Environ. Sci. Technol. 46 (4), 2018–2024.

H. Zhong et al. / Energy Policy 83 (2015) 14–25

Hu, X., Murgovski, N., Johannesson, L., Egardt, B., 2013. Energy efficiency analysis of a series plug-in hybrid electric bus with different energy management strategies and battery sizes. Appl. Energy 111, 1001–1009. Yang, Z., Kang, L., Niu, Q., et al., 2014. A self-learning TLBO based dynamic economic/environmental dispatch considering multiple plug-in electric vehicle loads. J. Mod. Pow. Syst. Cle. Energ. 2 (4), 298–307. Ocran, T., Cao, J., Cao, B., et al., 2005. Artificial Neural Network Maximum Power Point Tracker for Solar Electric Vehicle. Tsinghua Sci. Tech. 10 (2), 204–208. Ding, Y., Yang, H., 2013. Promoting energy-saving and environmentally friendly generation dispatching model in China: phase development and case studies. Energy Policy 57, 109–118. NDRC, MEP, SERC, and NEA, 2007. Measures for Energy-Saving Generation Dispatching (Interim). Central Government of PRC, Beijing, No. 53 document. Gao, C., Li, Y., 2010. Evolution of China's power dispatch principle and the new energy saving power dispatch policy. Energy Policy 38, 7346–7357. Xu, G., Kang, C., and Xia, Q., 2009. Research on market benefit evaluation and competition mode of energy-conservation based generation dispatch. In: Proceedings of the CIGRE Guilin Symposium. Liu, T., Yang, Y., Li, X., et al., 2010. Energy saving generation dispatch for sustainable development of energy utilizing in china. In: Proceedings of the Power and Energy Engineering Conference, APPEEC, Asia-Pacific pp. 1–4. China Electricity Council, 2013b. Statistics Express of China’s Electric Power Industry (2012),〈http://www.cec.org.cn/guihuayutongji/tongjxinxi/niandushuju/ 2013-11-07/111737.html〉, (accessed 13.09.14.). Wood, A.J., Wollenberg, B.F., 1996. Power Generation, Operation, and Control, second ed. John Wiley and Sons, New York, United States of America. Bard, J.F., 1988. Short-term scheduling of thermal-electric generators using Lagrangian relaxation. Oper. Res. 36, 756–766. Shahidehpour, M., Yamin, H., Li, Z., 2002. Market Operations in Electric Power Systems. Wiley, New York, United States of America. Fu, Y., Shahidehpour, M., 2007. Fast SCUC for large-scale power systems. IEEE Trans. Power Syst. 4, 2144–2151. Vargas, L.S., Quintana, V.H., Vannelli, A., 1993. A tutorial description of an interior point method and its applications to security-constrained economic dispatch.

25

IEEE Trans. Power Syst. 8, 1315–1324. Carrion, M., Arroyo, J.M., 2006. A computationally efficient mixed-integer linear formulation for the thermal unit commitment problem. IEEE Trans. Power Syst. 3, 1371–1378. Arroyo, J.M., Conejo, A.J., 2000. Optimal response of a thermal unit to an electricity spot market. IEEE Trans. Power Syst. 3, 1098–1104. Arroyo, J.M., Conejo, A.J., 2004. Modeling of start-up and shut-down power trajectories of thermal units. IEEE Trans. Power Syst. 3, 1562–1568. Biskas, P.N., Bakirtzis, A.G., Macheras, N.I., et al., 2005. A decentralized implementation of DC optimal power flow on a network of computers. IEEE Trans. Power Syst. 1, 25–33. Ng, W.Y., 1981. Generalized generation distribution factors for power system security evaluations. IEEE Trans. Power Appar. Syst. 100, 1001–1005. Fletcher, R., 1987. Practical Methods of Optimization, 2nd ed. Wiley, New York, US, pp. 357–416. Zimmerman, R.D., Murillo-Sánchez, X., Nchez, C.E., et al., 2011. Matpower: steadystate operations, planning, and analysis tools for power systems research and education. IEEE Trans. Power Syst. 26, 12–19. Wang, S.J., Shahidehpour, S.M., Kirschen, D.S., et al., 1995. Short-term generation scheduling with transmission and environmental constraints using an augmented Lagrangian relaxation. IEEE Trans. Power Syst. 3, 1294–1301. Xie, L., Ilic, M. D., 2010. Emission-concerned economic dispatch: possible formulations and implementations. In: Proceedings of the IEEE PES Transmission and Distribution Conference and Exposition, pp. 1–6. The IBM CPLEX Optimizer, 2014.〈http://www-01.ibm.com/software/commerce/op timization/cplex-optimizer/index.html〉(accessed 28.08.14.). Gurobi Optimizer, 2014.〈http://www.gurobi.com〉(accessed 28.08.14.). Bohai-rim Steam-Coal Price Index (BSPI), 2012.〈http://osc.cqcoal.com/news/V02/ 57921_1.html〉(accessed 28.01.15.). Chinese National Standard, 2008. Generic Computational Rules for Composite Energy Consumption. GBT 2589—2008.