Electricity Generation and Water- Related Constraints: An Empirical Analysis of Four Southeastern States Gbadebo Oladosu, Stan Hadley, D.P. Vogt, and T.J. Wilbanks Oak Ridge National Laboratory, Oak Ridge Tennessee September 26
Introduction Water-Energy Nexus Concerns Increasing Attention: Assessments, Workshops and Reports 25 Energy Policy Act Section Premise Energy-Water Nexus Initiative in the DOE National Laboratory Team Roadmapping and Report to Congress undergoing Review Many Dimensions: Water-Energy; Energy-Water Water and Energy are both Indispensable Anecdotal Evidence California 2/21 26 Drought Plant Siting Decisions Water-Related Regulations
Introduction Crucial Analytical Need Lack of Quantitative Estimates of Impacts Past Projections of Water Use have been inaccurate More Comprehensive Approaches required Objectives: Focus on Electricity Supply Largest water withdrawals Sensitive on both demand and supply side to Climate/Hydrological changes Employ empirical data to attempt quantification of water availability constraints on Electricity Supply Network Properties of both Water and Electricity Crucial
Introduction Prototype Work to Integrate Hydrologic and Economic Data for Water-Energy Nexus Analysis Regional Precipitation and Stream Flow Network Regional Economy County Water Availability Indicator Electricity Sector
Introduction
Regional Water Resources Alabama, Florida, Georgia and Tennessee Abundant Aggregate Water Supply Normal Annual Rainfall: >5 inches Tennessee River is the Largest Tributary of the Ohio River which empties into the Mississippi Water Issues Considerable Sub-Regional Variation Drought periods have become more frequent and longer Dispute among Alabama, Georgia and Florida Tennessee Inter-basin Water Transfer Act
Regional Electricity Sector Electricity Supply by Regulated Utilities 3 Large Utilities (1 Public, 2 Investor-Owned) Other Public Utilities Independent Producers Intimate Relationship between Water and Electricity Supply Surface resources used to meet most cooling water needs TVA: Manages 65-mile long Tennessee River Basin TVA: Largest Public Power Company in the United States (8 Million Customers)
Note: Other Cooling includes Gas Turbines, Other Cooling Systems (Dry Cooling), and Plants between 1-1MW for which detailed cooling data are not collected. In addition, starting from 21 no data are collected for Nuclear Plants. Regional Electricity Sector Distribution of Net Generation by Company (24) Distribution of Net Generation by Fuel (24) (Percent) 1 8 6 4 2 Alabama Florida Georgia Tennessee Total Alabama Electric Coop Inc Alabama Pow er Co* Florida Pow er & Light Co Florida Pow er Corp Georgia Pow er Co* Gulf Pow er Company* JEA Southern Pow er Tennessee Valley Authority Others (Percent) 7 6 5 4 3 2 1 Alabama Florida Georgia Tennessee Total Coal Petroleum Natural Gas Nuclear Hydropower Others Distribution of Capacity by Cooling System (23) Distribution of Capacity by Prime Mover (23) (GW) 6 5 4 3 2 1 Alabama Florida Georgia Tennessee (GW) 6 5 4 3 2 1 Alabama Florida Georgia Tennessee O-Fresh Water O-Saline Water R-Pond/Canal R-Forced R-Induced R-Natural Hydropower Other Hydro (Conventional) Hydro (Pumped) Steam Turbine Gas Turbine CC-Steam Turbine CC-Gas Turbine ICE Others
Modeling Approach Short-Run Approach Slow Adjustment of Capital Non-Storable Good: Continuous Supply-Demand Balance Non-Price Related Daily Fluctuations in Demand Can provide insight into long-run decision variables Electricity Sector Modeling in the Literature Programming Models: Disaggregate; Step Cost Functions Econometric Models: More Aggregate, Smooth Cost Functions
Modeling Approach Current Study: Characteristics-based Hybrid Approach Plant Level: Defined by EIA Utility and Plant Id Monthly: Available Data Plant Contribution to Meeting Demand determined by: Fuel/Electricity Prices Capacity Fuel Type(s) Prime-mover Type(s) Cooling Technology Type(s) Water Supply Indicator
WS p,m : Monthly Water Supply Indicator at Plant County Location; D p,c, D p,f, and D p,g are distribution of plant capacity by cooling technology, fuel types, and prime mover respectively Modeling Approach Model Equations: log q p, m log = log K + + β f p β wr, m + β f, m k, m D p, f logwr log K p + + p c + β β u c, m β k 2, m D u, m d p, c log K p, u + β log p hws, m.log K WS p p, m + + β cws, m ( Ps, m / UC p, m ) + ε p, m ( D. P ) UC p, m = α m + β uc, m f p, f log f, m G = m p q p, m q p, m = q p, m + ε p, m g β g, m D p, g ( logwr ) p / WS p, m G m : Monthly Regional Net Generation; q p,m : Monthly Plant Generation; K p : Plant Capacity; P s,m : Monthly Average State Electricity Price; P s,f,m : Monthly Average State Fuel Price; WR p : Plant Design Water Use Rate;
Model Estimation: Maximum Entropy Approach Example: Regression Model : y i = β k X i, k + ε i k Based on work popularized by Judge and Golan, which in turn is derived from the information theory work of Shannon and Jaynes (Fraser, 2). The entropy approach is based on the assumption that information contained in an observation of a random variable s s value (or equivalently the level of uncertainty about a possible value) is inversely proportional to its probability. The negative of the log of the probability has been shown to be an optimal measure of that information (information score). Shannon s s total entropy measure for a random variable is the probability weighted sum of the information scores of each possible value. This is maximized when each p n equals 1/n (Complete Uncertainty). Based on Jayne s s proposal the Entropy Approach to estimation proceeds by formulating an entropy objective function that is maximized
Model Estimation: Maximum Entropy Approach Example: Regression Model : Maximize: Subject to: H ( p n, k, vm, i ) = pn, k log pn, k vm, i log n, k m, k y i = k β k X i, k + ε i v m, i p n,k and v m,i are probabilities associated with parameters and residual terms, respectively β k = ε = i n p n k m v m i n p n, k z n, k m v m, i w m, i, = 1, = 1 z n,k and w m,i are known as support points contained within an interval wide enough to contain the true parameter values and residuals, respectively.
Model Estimation: Maximum Entropy Approach The modeler provides three types of information, with implications for estimated coefficients. These are: Weight on each information component (i.e. parameters and residual) in the objective; Choice of the number of support points n and m; and Boundaries on the parameter and error support points Recommendation is that n and m are greater than 2. Its being shown that values of greater than 3 do not improve the solution significantly, while considerably increasing its size. We use a value of 3. Without prior information, boundaries for the parameters should be wide and centered at zero. We use (-1,,1). Golan and Judge recommend that error term boundaries be set at 3 times the standard deviation (σ) of the dependent variable, but may be increased to ensure feasibility. We use (-5σ,, -5σ)
arameter Estimates Months 1 2 3 4 5 6 7 8 9 1 11 12 Economic Coefficients β uc,m : Cost Parameter 2.238 2.456 3.464 5.549 5.411 -.132 -.312 -.387 1.233 -.47 2.786.753 α m : Cost Constant..411-2.551-1.172-1.448 2.26 2.141 2.146.131 2.554-2.675.984 β u,m : Markup Parameter (Utility Size <= 1GW).94.8 -.55.165.175 -.14 2.798 1.883-1.822 2.744-1.587-1.316 β u,m : Markup Parameter (Utility Size >1 and <= 5GW).191.274.99.574.419-4.243-4.725-4.193 -.129 -.241.239 1.33 β u,m : Markup Parameter (Utility Size >5 and <= 1GW).55.91.116.152.159 -.759.765.714 -.249 -.364.22 1.25 β u,m : Markup Parameter (Utility Size > 1GW).137.166.125.224.172-1.381 -.67 -.122 -.18 -.34.235 1.257 Plant Capacity Coefficients β k,m : Plant capacity: -.182 -.138 -.139 -.321 -.247 -.269 -.286 -.272 -.15 -.27 -.314 -.263 β k2,m : Square of Plant Capacity ():.8.5.6.19.13.16.18.17.5.11.14.12 Prime Mover Coefficients (Base: Steam Trubine) β g,m : Hydro (Conventional) -.659-1.889-1.688-2.197-1.145-1.316 -.961 -.749-1.239 -.965 -.911 -.416 β g,m : Hydro (Pumped) -2.119-3.5-1.669-3.381-1.87-1.993 -.917-1.293.4.123.3 -.165 β g,m : Gas Turbine -.742 -.64 -.611 -.363 -.656-1.63 -.833 -.853-1.27 -.633 -.211 -.65 β g,m : Combined-Cycle - Steam Turbine -1.228-1.332 -.878.995.153.38 -.56.175.22.491 -.29 -.166 β g,m : Combined-Cycle - Gas Turbine -.232 -.187.561.84.48.345.1.264 1.457 1.71 2.847 1.65 β g,m : Internal Combustion Engine -.257.26 -.921 -.556 -.32.432.367.44 1.91 -.179 1.247 -.837 Fuel Type Coefficients (Base: Coal) β f,m : Petroleum -1.77 -.932 -.764.51.158 -.254 -.499 -.566 -.546 -.91-1.482 -.791 β f,m : Natural Gas -.399 -.36 -.81.868.831.169 -.54.83 -.249 -.445 -.38.411 β f,m : Nuclear -.116 -.135 -.388-1.131 -.967 -.16.15.133.341 -.231 -.548 -.726 β f,m : Others.921 1.273.32-2.534-2.348-1.384-1.258-1.645-1.61-2.827-3.83-3.422 Cooling System Coefficients (Base: O-Fresh Water) β c,m : Once-through Saline Water -.21.164.254.137.51.213 -.78.27.59.284.18 -.323 β c,m : Recirculating with Pond/Canal -.37.172.182.213.439.44.93.128.33.434.858.193 β c,m : Recirculating with Forced Draft Cooling Tower -.597 -.845 -.64 -.512. -.52 -.237 -.232 -.388 -.359 -.3 -.637 β c,m : Recirculating with Induced Draft Cooling Tower -.481 -.344 -.236.191 -.149.21 -.121 -.154 -.96 -.25 -.386 -.52 β c,m : Recirculating with Natural Draft Cooling Tower.54.15 -.154 -.61.122 -.27 -.18 -.144.31 -.62.91 -.2 β c,m : Others (Except Hydro).245.76 -.423 -.429.27 -.156 -.116 -.29 -.18 -.77.232 -.69 Water Use Coefficients β wr,m : Cooling Water Rate.52 -.1 -.57 -.36 -.35 -.56 -.23 -.32 -.15 -.4 -.6 -.14 β hws,m : Water Availability Measure (Hydro).299 1.347.376 1.541..46.... 1.829. β cws,m : Cooling Water Rate-Water Availability Interaction -.41 -.2. -.18........
arameter Estimates: T-Statistics Months 1 2 3 4 5 6 7 8 9 1 11 12 Economic Coefficients β uc,m : Cost Parameter 24.5 52.1.7 1.9 1.5-4. -3.5-5.7 1.8-4.6 1.3 7.9 α m : Cost Constant. 17. -2.3 -.7 -.4.7 42.6 66.1.4 41.2 -.9 6.9 β u,m : Markup Parameter (Utility Size <= 1GW).3. -.2 1.3.8. 4.6 4.1-5.9 2.9-2.1-2.5 β u,m : Markup Parameter (Utility Size >1 and <= 5GW) 1.7 1.1.5 1.7 1.9-37.8-7.4-7. -.3 -.4 3.3 9.7 β u,m : Markup Parameter (Utility Size >5 and <= 1GW).7.6 1.3 1.7 1. -7.2 5.1 4.8-2.1-2.1 1.7 8.8 β u,m : Markup Parameter (Utility Size > 1GW) 1.1 1.4 1.1 1.7 1. -13.6-1.1-1.8-3.6-2.5 1.5 8.3 Plant Capacity Coefficients β k,m : Plant capacity: -6.1-4.3-3.1-7.5-5.7 -.4-9.9-1.3-3. -5. -7.3-7.1 β k2,m : Square of Plant Capacity (): 3.7 2.1 2.1 6.5 4.4.6 9. 9.2 2.1 3.8 4.9 4.8 Prime Mover Coefficients (Base: Steam Trubine) β g,m : Hydro (Conventional) -1. -4.2-2.2-1.5-1.6-1. -1.8-1.4-1.5-2.3-1.8-1.4 β g,m : Hydro (Pumped) -1.5-2.3 -.9-1.3 -.4 -.4 -.5 -.8.... β g,m : Gas Turbine -4.7-3.5-3.3-2.1-3.6-8. -7.3-8.1-7.4-3.4-1.1 -.3 β g,m : Combined-Cycle - Steam Turbine -1.6-1.6 -.4 1.8.3 1. -.2.6.1 1.1 -.1 -.4 β g,m : Combined-Cycle - Gas Turbine -.7 -.6 1.5 2..1 1.3.4 1.2 5.6 4.6 7.8 2.6 β g,m : Internal Combustion Engine -.1. -.1 -.1 -.2.6.5.6 1.3 -.1.8-1. Fuel Type Coefficients (Base: Coal) β f,m : Petroleum -9.6-8.1-5.7.3 1.1-2.7-6.7-7.6-5.5-6.6-8.7-4.6 β f,m : Natural Gas -3.1-2.1 -.5 4.8 4.5 1.3 -.5.8-1.8-2.4-1.5 2. β f,m : Nuclear -2.2-3.2-5. -1.3-9.3-2.2 2.3 3.2 5.8-3.3-6.8-9.5 β f,m : Others 1.2 1.4.1-4. -3.9-3.4-3.2-4.6-3.5-4.4-5.9-4.8 Cooling System Coefficients (Base: O-Fresh Water) β c,m : Once-through Saline Water -.5 3.5 4.9 2.6.8 4.8-1.9.7 1.3 4. 2.6-5.7 β c,m : Recirculating with Pond/Canal -.7 3.1 2.3 3.4 5.1 7.9 1.7 2.4.5 5.1 11.5 3. β c,m : Recirculating with Forced Draft Cooling Tower -5.5-6.1-4.5-3.4. -.6-2.8-2.9-3.4-2.6 -.2-5. β c,m : Recirculating with Induced Draft Cooling Tower -4.2-3.1-2.4 2.1-1.3.3-1.8-2.3-1.1-2.2-2.8 -.6 β c,m : Recirculating with Natural Draft Cooling Tower.8.2-2.2 -.8 1.7 -.5-3.8-3.2.5 -.9 1.4 -.3 β c,m : Others (Except Hydro) 2.3.6-3.7-3.7.2-1.9-1.6-3. -1.9 -.7 2.5 -.8 Water Use Coefficients β wr,m : Cooling Water Rate 4. -.1-4.4-2.7-2.7-5.4-2.6-3.8-1.4-3.2 -.5-1.3 β hws,m : Water Availability Measure (Hydro).5 4.3.8 1...1.... 3.2. β cws,m : Cooling Water Rate-Water Availability Interaction -8.7-7.2. -22.4 -.2.......
arameter Estimates: Cooling-Related Once-through Saline Water Recirculating with Pond/Canal Recirculating with Forced Draft Cooling Tower Recirculating with Induced Draft Cooling Tower January -.21 -.37 -.597 -.481.54.245.52.299 -.41 February.164.172 -.845 -.344.15.76 -.1 1.347 -.2 March.254.182 -.64 -.236 -.154 -.423 -.57.376. April.137.213 -.512.191 -.61 -.429 -.36 1.541 -.18 May.51.439. -.149.122.27 -.35.. June.213.44 -.52.21 -.27 -.156 -.56.46. July -.78.93 -.237 -.121 -.18 -.116 -.23.. August.27.128 -.232 -.154 -.144 -.29 -.32.. September.59.33 -.388 -.96.31 -.18 -.15.. October.284.434 -.359 -.25 -.62 -.77 -.4.. November.18.858 -.3 -.386.91.232 -.6 1.829. December -.323.193 -.637 -.52 -.2 -.69 -.14.. Recirculating with Natural Draft Cooling Tower Others (Except Hydro) Cooling Water Rate Water Availability Measure (Hydro) Cooling Water Rate-Water Availability Interaction Cooling Water Use Rate Coefficients are (mostly) Negative as Expected Energy Penalty Effect seen in the Coefficient of Closed Cycle Systems Although bounded to avoid incorrect signs, water supply indicator coefficients are non-zero for several months
Ex-post Predictive Performance In-Sample Prediction (1996) y =.978x R 2 =.8624 3 25 2 15 1 5 5 1 15 2 25 Actual Net Generation: GWh) Prediction Linear (Prediction) Out-of-Sample Prediction (1998) y =.916x R 2 =.8138 3 25 2 15 1 5 5 1 15 2 25 Actual Net Generation: GWh) Prediction Linear (Prediction) Out-of-Sample Prediction (1997) y =.93x R 2 =.8192 3 25 2 15 1 5 5 1 15 2 25 Actual Net Generation: GWh) Prediction Linear (Prediction) Out-of-Sample Prediction (1999) y =.9197x R 2 =.8387 3 25 2 15 1 5 5 1 15 2 25 Actual Net Generation: GWh) 25 2 15 1 5 25 2 15 1 5 In-Sample Prediction (2) y =.8891x R 2 =.8791 5 1 15 2 25 Actual Net Generation: GWh) Prediction Linear (Prediction) Out-of-Sample Prediction (22)y =.8916x R 2 =.875 5 1 15 2 25 Actual Net Generation: GWh) Prediction Linear (Prediction) Prediction Linear (Prediction)
Simulations: Calibrated to 22 Data Scenario 1: Change in Net Generation under a 1% Reduction in the Water Availability Indicator (Unconstrained by Demand) -.5-1 -1.5-2 -2.5-3 January February March April May June July August September October November December January February March April May June July August September October November December January February March April May June July August September October November December January February March April May June July August September October November December (Percent) Alabama Florida Georgia Tennessee
Simulations: Calibrated to 22 Data Scenario 2: Change in Net Generation under a Replacement of All Once-Through with Recirculating Natural Draft Cooling Systems (Unconstrained by Demand) 1 8 6 4 2-2 -4-6 -8-1 -12 January February March April May June July August September October November December January February March April May June July August September October November December January February March April May June July August September October November December January February March April May June July August September October November December (Percent) Alabama Florida Georgia Tennessee
Simulations: Calibrated to 22 Data Scenario 3: Change in Net Generation under a 1% Reduction in Cooling Water Use Rate (Unconstrained by Demand).5.4.3.2.1 -.1 -.2 January February March April May June July August September October November December January February March April May June July August September October November December January February March April May June July August September October November December January February March April May June July August September October November December (Percent) Alabama Florida Georgia Tennessee
Simulations: Calibrated to 22 Data State-Level Summary of Net Generation Changes under Three Scenarios (Unconstrained by Demand) (Percent).4.2 -.2 -.4 -.6 -.8-1 -1.2-1.4-1.6 Alabama Florida Georgia Tennessee Total Scenario 1: 1% Reduction in Water Availability Indicator Scenario 2: Replacement of All Once-Through with Recirculating Natural Draft Cooling Systems Scenario 3: 1% Reduction in Cooling Water Use Rate
Conclusion Approach incorporates advantage of programming and econometric methods Model Performs well In- and Out-of Sample Stable Operational Structure (?) Water-Related Coefficients are Significant Simulations suggest sizeable impacts on plant operations Cooling Technologies Water Availability Design Water Use Rates
Conclusion Further Work Data Issues Water Use Rates and Cooling Types missing for a significant amount of capacity in each state Periodicity of Data Inadequate Use a full Hydrological model to provide Water Availability Data Incorporate Pollution Equipment Data Include an Explicit Cost Function as part of model More Detail Generator Level (?) Explicit Transmission/Distribution Constraints Case Studies at Sample of Plants Modeling for Less Water-Endowed Regions Modeling for More Competitive Electricity Markets Consider how to include other features of electricity water