Copyright 2009 by all authors. Published and used by MIT ESD and CESUN with permission.

Size: px

Start display at page:

Sydney Robertson
5 years ago
Views:

1 Second International Symposium on Engineering Systems MIT, Cambridge, Massachusetts, June 15-17, 2009 A Cost-Effectiveness Analysis of Alternative Ozone Control Strategies: Flexible Nitrogen Oxide (NOx) Abatement from Power Plants in the Eastern United States Lin Sun 1, Mort Webster 2*, Massachusetts Institute of Technology, Cambridge, MA Gary McGaughey 3, Elena McDonald-Buller 4, Tammy Thompson 5 and David Allen 6 University of Texas, Austin, TX Copyright 2009 by all authors. Published and used by MIT ESD and CESUN with permission. ABSTRACT The dependence of ozone formation on meteorology makes the timing of emissions important, suggesting the need for a temporally differentiated regulation for the emissions of its precursor: nitrogen oxide (NOx). Such a flexible NOx regulation policy, so-called smart trading, which is designed to target ozone episodes by reducing NOx emissions prior to and during forecasted episodes, has the potential for reducing the compliance cost and helping to solve the persistent ozone non-attainment problem in the Eastern U.S. However, the total compliance cost of this new policy depends critically on the accuracy of ozone forecasting. This study aims to address the question of whether a NOx trading program that differentiates across emissions by time could reduce ozone concentrations more cost-effectively than the conventional command-andcontrol method in the Eastern United States given the uncertainty in ozone forecasting. A costeffectiveness analysis is conducted to compare the average cost for achieving a certain amount of ozone reduction under the proposed smart trading plan and the command-and-control policy. The probability distribution of the compliance cost under a smart trading policy is simulated using a stochastic decision model based on the simulated electricity generation and air quality data. This study demonstrates the scientific and economic feasibility of a time-differentiated trading scheme and justified such a regulation through sensitivity analysis of ozone prediction errors. Keywords: flexible NOx regulation, ozone non-attainment, time-differentiated cap-and-trade, stochastic decision analysis, cost-effectiveness analysis, engineering systems. 1.Ph. D. Graduate, Department of Earth, Atmosphere and Planetary Sciences (EAPS) and Engineering System Division (ESD), linsun@mit.edu 2. Assistant Professor, ESD, mort@mit.edu 3. Research Scientist, Center for Energy and Environmental Resources (CEER), garym@mail.utexas.edu 4. Research Associate Professor, CEER, ecmb@mail.utexas.edu 5. Postdoctoral Researcher CEER, ftammv@.cne.utexas.edu' 6. Professor, CEER, allen@che.utexas.edu

2 1. Introduction It has been nearly four decades since the passage of the Clean Air Act and the establishment of the National Ambient Air Quality Standard (NAAQS) for ozone and other criteria pollutants. Emission cap-and-trade programs have drawn great attention over the past two decades by promising an economically efficient mechanism for reducing air pollution [1]. The East Coast s NOx Budget Program since 1999 has lead to considerable reduction in NOx emissions from point sources over the years. But due to complexities of implementation and high costs of emissions reductions many regions of the Eastern U.S. remain in non-attainment of the ozone standard. As easier, less expensive emissions reductions are made, and additional permanent or annual reductions have even higher marginal costs, it becomes worthwhile to consider more costeffective approaches to control ozone levels, such as differentiated cap-and-trade programs. If options are restricted to costly technological mandates or further reduction in the seasonal cap, further NOx reductions may become politically infeasible, and ozone levels that harm human health could be allowed to persist as they have already for decades. One potential reason for the persistency is that ozone formation depends not only on the quantities of precursor (NOx) emissions but on their timing and location. Past research has revealed that ozone formed by the same amount of NOx could vary by a factor of five under different meteorological conditions [2]. This provides the scientific justification for a time- and location-differentiated NOx regulation. Montgomery [3] theoretically proved the existence of the market for an emission-trading program that creates pollution licenses, which conferred the right to emit until the emissions had a given level of impact on a set of air pollution monitors. He also showed the advantages of pollution licenses over emissions licenses in terms of applicability. Mendelsohn [4] showed that the variations in the environmental impacts of pollutant emissions would result in welfare losses if emissions were treated homogeneously. In addition, Tietenburg [5] found empirical evidence of above problems by showing that undifferentiated permit-trading programs sometimes required a higher degree of control and thus higher costs to achieve the same improvements in air quality than prescriptive regulations. All of these indicate the validity and necessity of a flexible NOx regulation that considers the time and location of the emissions. As the first step, this study focuses on the time-differentiated regulation only. Smart trading is one example of flexible NOx regulation that is potentially more cost-effective in reducing ozone concentrations, since it targets lowering NOx emissions from electrical power plants only on ozone episode days. Hypothetically, the price of NOx allowance could be determined by the local administrative agency one day in advance, so that NOx polluters could adjust their production for the next day. One way to achieve short-term NOx emissions reductions from power plants is through electricity re-dispatch, shifting some of the electricity generation to units with lowest NOx emission rates. There are other ways to achieve short-term NOx reductions, such as operational modification. However, the incremental costs associated with the operational changes are difficult to estimate so that these responses are not modeled. The NOx emission changes under electricity redispatch thus provides a conservative estimate of the cost-effectiveness of a time-differentiated regulation for NOx emissions from point sources. One of the major technical barriers to a time-differentiated approach is the necessary reliance on weather and atmospheric chemistry forecasting, which is generally perceived to have large

3 uncertainty. The transaction costs of a time-differentiated program would be too high if ozone forecasting is not sufficiently accurate. Specifically, false positive events (a forecasted high ozone day turns out to be a low ozone day) increase the unnecessary cost on non-ozone days; while false negative events (a forecasted low ozone day turns out to be a high ozone day) decrease the amount of ozone that could otherwise be reduced. The costs of unnecessary reductions and the costs of failing to comply with the NAAQS must be included into the decision analysis when evaluating this new policy. It is important to conduct cost-effectiveness analysis for the proposed smart trading policy. Smart trading does not require additional NOx control technology; rather, it provides flexibility to power plant operators to decide whether it is economically preferable to reduce electricity generation to reduce NOx emissions or to generate electricity, and, if necessary, buy NOx permits from other plants. In this way, the NOx emissions and ozone concentrations in the whole area are reduced with lower cost than installing NOx control technologies such as Selective Catalytic Reduction (SCR) and Selective Non-Catalytic Reduction (SNCR), while still satisfying the total electricity demand in the area. Over time, if the cost of reducing NOx emissions through dispatching electricity generation within an electrical network becomes too expensive, NOx control technologies could become a cost-effective strategy for some generators. In other words, smart trading can also provide an incentive for power producers to adopt NOx control technologies (this multi-year decision making is not modeled here). As a first step, it is essential to compare the cost of reducing NOx emissions and ozone concentrations under both smart trading policy designs and under a pollution control technology policy design in order to determine the circumstances under which each policy design would out-perform the other. This study aims to evaluate whether current weather and air quality forecasting abilities is within the needed minimum threshold of accuracy. If it is, further investigation of this regulatory approach is warranted. If not, this analysis could also provide guidelines for the EPA to estimate the potential value of improving their models, improving emission inventories, and increasing computational capacity. 2. Methods 2.A Simulation of the Electrical power system The Classic PJM power system is the original Pennsylvania, New Jersey, Maryland power system that also includes the District of Columbia and Delaware. PJM is used here to analyze the potential behavior of a typical electrical power system under the proposed time-differentiated NOx cap-and-trade program in the Eastern U.S.. The electricity generation re-dispatch under different NOx prices is modeled using the PowerWorld Simulator under the Optimal Power Flow (OPF) mode. The PJM Financial Transmission Rights (FTR) base-case power flows is used to simulate Classic PJM [6]. NOx emissions and NOx allowance prices are incorporated into the variable costs of the power plants as an additional fuel cost. The cost curves are generated as described in [6]. The NOx price is applied uniformly to all units in PJM with $2k/ton reproduces the policy as of 2002 (defined as the base case). Smart trading policy scenarios within the context of this study consist of increasing the base case NOx price to $30k, 50k, 100k/ton.

4 Most NOx emissions in the Eastern U.S. are from coal-fired plants; and most of these coal-fired plants have already installed primary control technologies [7]. Thus secondary NOx control technologies are expected to play an important role in controlling point source NOx emissions in the Eastern U.S.. SNCR is one of the most discussed secondary NOx control technologies. Therefore, this study compares the cost-effectiveness of smart trading under different NOx prices with that of installing SNCRs to all coal-fired power plants in PJM. 2.B Simulation of Ozone Impacts of NOx emissions Changes The Comprehensive Air Quality Model with Extensions (CAMx) is used to perform the photochemical grid modeling for this project [8]. This analysis uses historical emission inventories of ozone precursors and meteorological data to simulate the hourly ground level ozone concentrations in the PJM area during June~September 2002 to represent a typical summer ozone season. The simulated daily maximum 8-h ozone concentrations for the grid cells where the 37 EPA ozone monitoring sites (from EPA AIRSNOW) during summer 2002 in the Philadelphia/Baltimore region are used for the cost-effectiveness analysis. 2.C Stochastic Decision Analysis In this study, an analysis tool is developed to integrate electrical power system and photochemical modeling into a stochastic decision analysis model. In the design of smart trading, the system operator needs to make the decision about the next day s NOx price based on the forecasted ozone level for the next day. If a high ozone level 1 is forecasted, a high NOx price will be announced for the next day. If the next day turns out to be a low ozone day, the decision to raise the NOx price will lead to unnecessary costs. Similarly, if a low ozone level is forecasted, the NOx price will not be raised. If the next day turns out to be a high ozone day, the decision to maintain the default NOx price will lead to continued ozone non-compliance, which could have been reduced if the forecast had been accurate (see Figure 1). The utility function for the decision model is defined as the ratio of the incremental costs (Δcost) to the incremental reductions of daily maximum ozone concentrations at the 37 ozone monitoring sites in the Philadelphia/ Baltimore region (ΔO 3 ), relative to the base case. ΔO 3 will be positive unless the forecast is a false alarm. In the case of forecast low ozone days, the base case NOx price is maintained, so that Δcost and ΔO 3 are, by definition, 0. Figure 1: Framework for deciding whether to choose smart trading as the ozone control policy in the presence of risks in ozone forecasts during an N-day period. H (L) represents a high (low) ozone day. 1 If the average 8-h ozone concentration across the 37 ozone monitoring sites is over 85 ppb, a day is considered as a high ozone day; and vice versa.

5 Assume U i and D i are the increased cost of electricity generation under a higher NOx price on the i th day of the N-stage decision-making process and the associated reduction in the daily maximum 8-h ozone concentration, respectively. The mean and probability distribution of Δcost/ΔO 3 associated with each smart trading case can be simulated by conducting the above decision analysis for a large number of sample days; In this analysis, a sample size of N=50,000 days is used. Denote the mean Δcost and ΔO 3 generated by the Monte Carlo method are U 1 and D 1, respectively, and the Δcost and ΔO 3 of the SNCR case as U 2 and D 2, respectively. Because U 2 and D 2 are not a function of p and q, the values for U 1 /D 1 obtained by the stochastic decision model will be used to compare to the deterministic U 2 /D 2. The rule for accepting smart trading is defined as: Δcost/ΔO 3 (smart trading)> Δcost/ΔO 3 (SNCR), or accept smart trading if U 1 /D 1 <U 2 /D 2, and reject smart trading otherwise. 3. Results 3.A Cost Analysis The incremental costs of NOx emissions reductions below the base case ($2k / ton) are determined for the SNCR case as the levelized costs for the capital expenditure plus variable operating costs, and as incremental generation costs at higher marginal cost units for the smart trading scenarios. Since the SNCR costs are a function of operation costs, SNCR is an effective control option for seasonal NOx reduction. The EPA published the seasonal SNCR cost for industrial boilers larger than 100 MMBtu/h (or 29 MW/h) to be $ /ton of NOx removed in year 1999 dollars, including capital and O&M costs [9]. We assume that SNCRs are installed at all coal-fired EGUs in PJM (98 units in total), most of which have capacities larger than 29 MW/h. The average daily NOx reduction as a result of installing SNCRs during the four-month summer season is tons, yielding an average daily SNCR cost to be $0.34 to 0.51 Million in year 1999 dollars. Since the daily costs for the smart trading are calculated in 2005 dollars, the SNCR cost is converted into 2005 real dollars in order to compare the cost of SNCR and smart trading. Assume the inflation rate is 5% from 1999 to 2005, the average daily SNCR cost during ozone season is then $0.46 to 0.69 Million, or $ Million in year 2005 dollars. The cost of smart trading is estimated using the increased variable cost of electricity generation under higher NOx price relative to the base price. 3.B Ozone Abatement Costs without Forecast Errors Assuming no ozone forecast errors, the abatement costs of smart trading under different NOx prices are simulated using the stochastic decision model. The average daily costs of smart trading and SNCR versus the average reduction in the daily 8-h maximum ozone concentrations over the 37 monitoring sites are plotted in Figure 2. Error bars (±σ) except the one on the SNCR cost represent uncertainty in the occurrence of high ozone events. These uncertainty ranges would narrow if the population size of the Markov model were increased by including additional ozone

6 season observations. The error bars on the SNCR cost represent uncertainty of the cost provided by the 2003 EPA fact sheet. As shown in Figure 2, the lines connecting 0, 30k, 50k and 100k cases form a cost-effectiveness frontier for ozone reductions. An ideal regulatory design would have zero cost and maximum ozone reductions, and would be located in the lower right corner. The frontier describes the trade-offs across the cost-effective options modeled; one can achieve greater ozone reductions, but only at an increased cost. Most notably is that the SNCR alternative is less cost-effective. Without ozone forecast error, one would almost never prefer the SNCR approach to any of the smart trading approaches Figure 2: Average daily cost of smart trading and SNCR versus average reduction of daily 8-h maximum ozone concentrations across over the 37 monitoring sites. 3.C Sensitivity Analysis of Forecast Errors The increased cost for reducing one ppb of daily 8-h maximum ozone (Δcost/ΔO 3 ) is used as the criterion to compare the cost-effectiveness of smart trading with the SNCR case scenario. However, both Δcost and ΔO 3 decrease with the false negative rate (q) increases. Thus if we focus exclusively on the ratio Δcost/ΔO 3, we might reach misleading conclusion that smart trading is more cost-effective than the SNCR case even though both Δcost and ΔO 3 under smart trading are very small. Therefore, a modified decision rule is used to conduct the decision analysis here in which a smart trading scenario is considered more cost-effective than the SNCR scenario if it has lower Δcost/ΔO 3 than the SNCR scenario, while reducing comparable (at least 70% in this study) of ozone under the SNCR scenario. The mean values for the cost-effectiveness measure, Δcost/ΔO 3, under the four scenarios are plotted for different rates of false positive and false negative errors (only $50k case scenario is shown in Figure 3 to compare with the SNCR case). Different colors represent different values of Δcost/ΔO 3, in the unit of Million $/ppb. Darker colors indicate lower values of Δcost/ΔO 3, which are more cost-effective. Points with higher Δcost/ΔO 3 than the SNCR scenario are shown in white (Area A). The black line indicates the approximate frontier where Δcost/ΔO 3 are equal for SNCR and smart trading. Area B indicates that the smart trading scenario has higher Δcost/ΔO 3,

7 but reduces less than 70% of ozone than the SNCR scenario (thus is not considered environmentally effective). Area C indicates that the smart trading scenario has lower Δcost/ΔO 3, and reduces at least 70% of ozone than the SNCR scenario, and thus is considered as more costeffective than the SNCR scenario. The dotted line indicates the approximate boundary where ΔO 3 for the two case scenarios equals. For any NOx price level, the higher the false negative rate, the lower is the required rate of false positives for smart trading to still be preferred. Figure 3: Δcost/ΔO 3 (Million $/ppb) of smart trading versus false positive rate and false negative rate for the $50k relative to the $2k/ton NOx case. As one estimate of the accuracy in ozone forecasting, the observed daily maximum 8-h ozone concentrations at the 37 sites are compared with the simulated concentrations. The average rates of p and q are shown in asterisk in Figure 3. The ozone forecasting models used in Philadelphia by the EPA from 2006 to 2008 (June~September) have yielded six different combinations of p and q (obtained from EPA), as shown in triangle in Figure 3. Based on these results, smart trading at $50k/ton NOx prices could be more cost-effective than the SNCR case for some models, suggesting that uncertainty in ozone forecasting may not be a major limiting factor for the feasibility of a time-differentiated NOx cap-and-trade program.

8 4. Conclusions The results presented here strongly support the effectiveness of smart trading in reducing NOx and ozone concentrations compared with the technology based control strategies. The capability of reducing ozone concentration by redispatching is not shown to be limited by network constraints, congestion, and inflexibility of dispatching electricity generation during peak demand hours. It is also shown that the ozone forecasting accuracy that is required for smart trading at $50k/ton of NOx to have lower cost for reducing per unit of ozone than the SNCR case is within the range of available accuracy of ozone forecasting and modeling, strongly suggesting that uncertainty in ozone forecasting may not be a major limiting factor for the feasibility of a time-differentiated NOx cap-and-trade program. References [1] Tietenberg, T. H., 1998, "Ethical Influences on the Evolution of the US Tradable Permit Approach to Pollution Control." Ecological Economics 24(2-3): [2] Mauzerall, D.L., Sultan, B., Kim, N. and Bradford, D.F., NOx emissions from large point sources: variability in ozone production, resulting health damages and economic costs, Atmospheric Environment, 39, [3] Montgomery, D.W.,1972, Markets in Licenses and Efficient Pollution Control Programs, Journal of Economic Theory, 5; [4] Mendelsohn, R.,1986, Regulating Heterogeneous Emissions, Journal of Environmental Economics and Management, 13(4); [5] Tietenberg, T., 1995, Tradeable Permits for Pollution Control when Emission Location Matters: What have We Learned? Environmental and Resource Economics 5: [6] Martin, K., 2008, Implementing a differentiated cap-and-trade program for ozone: flexible nitrogen oxide abatement from power plants, Ph.D. Dissertation, Massachusetts Institute of Technology, Engineering System Division, Cambridge, MA. [7] NESCAUM, 1998, Status Report on NOx: Control Technologies and Cost Effectiveness for Utility Boilers Executive Summary, June 1998 Retrieved from [8] ENVIRON, User s Guide Comprehensive Air Quality Model with Extensions (CAMx) version 5.01, ENVIRON International Corporation. Novato, California. [9] U.S. EPA 2003, Air Pollution Control Technology Fact Sheet SNCR, EPA report no. EPA- 452/F , July, retrieved from

Lin Sun. B.S. Chemistry, Peking University, China, 2003 Master of Applied Statistics, Ohio State University, 2006

Lin Sun. B.S. Chemistry, Peking University, China, 2003 Master of Applied Statistics, Ohio State University, 2006 A Cost-Effectiveness Analysis of Alternative Ozone Control Strategies: Flexible Nitrogen Oxide (NOx) Abatement from Power Plants in the Eastern United States by Lin Sun B.S. Chemistry, Peking University,