Slide 1 The Renewables Challenge: Keeping the lights on while managing variability and uncertainty Princeton University Academic Mini-Reunion October 2,

The Renewables Challenge: Keeping the lights on while managing variability and uncertainty Princeton University Academic Mini-Reunion October 2, 2015 Warren B. Powell PENSA Laboratory Dept. of Operations Research and Financial Engineering http://energysystems.princeton.edu

Outline  The renewables challenge  Combining wind and solar  The uncertainties of energy  Three energy problems » An energy storage problem » An energy portfolio policy model » Simulating the PJM grid using SMART-ISO  Concluding remarks

Wind in the U.S.

99.9 percent from renewables! Fossil Backup Battery Storage Wind & Solar 20 GW 750 GWhr battery!

Outline  The renewables challenge  Combining wind and solar  The uncertainties of energy  Modeling sequential decision problems under uncertainty  Three energy problems » An energy storage problem » An energy portfolio policy model » Simulating the PJM grid using SMART-ISO  Concluding remarks

Wind energy in PJM  Total PJM load plus actual wind (July) 53 wind farms

Wind energy in PJM  Total PJM load plus actual wind (July) 101,000 MWhr battery $50 billion!! Wind ~ 37 percent of total load

Solar energy  Solar from all PSE&G solar farms

Solar energy  Total PJM load plus factored solar (July) Solar ~ 15 percent of total load

Combining wind and solar  Mixture of wind and solar to meet July load 815,000 MWhr battery $989 billion!!

Combining wind and solar  Mixture of wind and solar to meet July load 260,000 MWhr battery $130 billion!!

Capacity factor analysis  Computing the “capacity factor” Capacity Actual

Capacity factor analysis  Classical analysis of renewables »Multiply installed capacity by the capacity factor 1 MW solar panel Capacity factor of.25 Translates to 0.25 MW of generation »Now, treat the.25 MW as if it is a form of conventional generation. »This makes it possible to scale up renewables without regard to the challenges of variability and uncertainty. »Let’s try to do better.

Northeast Reliability Councils and Interconnects

Energy from wind 1 year  Wind power from all PJM wind farms Jan Feb March April May June July Aug Sept Oct Nov Dec

Energy from wind 30 days  Wind from all PJM wind farms

Modeling wind  Forecast vs. actual for a single wind farm Actual Forecasted

Solar energy  Princeton solar array

PSE&G solar farms  Solar output over entire year (all farms) Sept Oct Nov Dec Jan Feb March April May June July Aug

Solar from PSE&G solar farms  Solar from a single solar farm

Solar from PSE&G solar farms  Within-day sample trajectories

Brazil

Rainfall  Foz do Iguaçu (Brazil) – 2011 through 2013 2011 2012 2013

Commodity prices  The price of natural gas »Reflects global and local economies, competing global commodities (primarily oil), policies (e.g. toward CO2), and technology (e.g. fracking).

$4 /mmBTU $120 /mmBTU!

Locational marginal prices on the grid LMPs – Locational marginal prices $58.47/MW

Locational marginal prices on the grid LMPs – Locational marginal prices $977/MW !!!

Locational marginal prices on the grid LMPs – Locational marginal prices $328/MW !

Locational marginal prices on the grid $52/MWhr

Uncertainty  It is important to separate: »Predictable variability Diurnal cycles Large weather patterns Major human events (Super bowl) »Stochastic uncertainty Temperature deviations from forecast Late/early arrival of a storm Generator failures Wind shifts PJM load Aggregate solar

Dealing with uncertainty Available at energysystems.princeton.edu

Policy function approximations  Battery arbitrage – When to charge, when to discharge, given volatile LMPs

 Grid operators require that batteries bid charge and discharge prices, an hour in advance.  We have to search for the best values for the policy parameters Policy function approximations

 Our policy function might be the parametric model (this is nonlinear in the parameters): Energy in storage: Price of electricity:

Policy function approximations  Finding the best policy »We need to maximize »We cannot compute the expectation, so we run simulations:

99.9 percent from renewables! »What answer do we get if we model this problem more carefully?

SMART-Invest  Features: »Finds the optimal mix of wind, solar and storage, in the presence of two types of fossil generation: Slow (steam) generation, which is planned 24 hours in advance Fast (turbine) generation, which is planned 1 hour in advance Real-time ramping of all fossils within ramping limits »Simulates entire year in hourly increments, to capture all forms of variability (except subhourly) »Minimizes investment and operating costs, possibly including SRECs and carbon tax. »Able to directly specify the cost of fossil generation (anticipating dramatic reduction in fossils). »Properly accounts for the marginal cost of each unit of investment.

SMART-Invest  The investment problem: Investment cost in wind, solar and storage. Capital investment cost in wind, solar and storage is the operating policy. Operating costs of fossil generators, energy losses from storage, misc. operating costs of renewables. Wind Solar

SMART-Invest Stage 1: Investment Hour 1Hour 2Hour 3Hr 8758Hr 8759Hr 8760 … Wind Capacity Solar Capacity Battery Capacity Fossil Capacity Wind Solar Find search direction Update investments

SMART-Invest  Operational planning »Meet demand while minimizing operating costs »Observe day-ahead notification requirements for generators »Includes reserve constraints to manage uncertainty »Meet aggregate ramping constraints (but does not schedule individual generators) 24 hour notification of steam 1 hour notification of gas Real-time storage and ramping decisions (in hourly increments) 36 hour planning horizon using forecasts of wind and solar

SMART-Invest  Operational planning model »Model plans using rolling 36 hour horizon »Steam plants are locked in 24 hours in advance »Gas turbines are decided 1 hour in advance Slow fossil running units 1 24 36 1 24 36 Slow fossil running units 1 24 36 Slow fossil running units 1 24 36 Slow fossil running units 1 24 36 Slow fossil running units 1 24 36 Slow fossil running units 1 24 36 Slow fossil running units 1 24 36 Slow fossil running units … … … 1 2345 8760 1 year Lock in steam generation decisions 24 hours in advance The tentative plan is discarded

SMART-Invest  Lookahead model with adjustment »Objective function »Reserve constraint: »Other constraints: Ramping Capacity constraints Conservation of flow in storage …. Tunable policy parameter

Robust policies  Policy search – Optimizing reserve parameter Low carbon tax, increased usage of slow fossil, requires higher reserve margin ~19 percent High carbon tax, shift from slow to fast fossil, requires minimal reserve margin ~1 pct

The value of storage  The marginal value of storage »On the margin, value of storage can be expensive! Energy in storage Time This investment in batteries is only used a small fraction of the time.

Policy studies  Renewables as a function of cost of fossil fuels Solar Wind $/MWhr cost of fossil fuels 100 80 60 40 20 0 Percent from renewable Total renewables 0 20 50 60 70 80 90 100 150 300 400 500 1000 3000 »Study assumes unconstrained access to wind at lowest cost (this is not available in the eastern U.S.) »Note the reluctance to introduce solar… »… and even greater reluctance to use storage. Battery

Policy studies  Sensitivity to CO2 tax. Slow fossil Fast fossil Nuclear Wind Solar “Other” fast fossil

The timing of decisions Day-ahead planning (slow – predominantly steam) Intermediate-term planning (fast – gas turbines) Real-time planning (economic dispatch)

The timing of decisions  The day-ahead unit commitment problem Midnight Noon Midnight Noon

The timing of decisions  Intermediate-term unit commitment problem 1:15 pm 1:45 pm 1:30 2:15 pm 1:00 pm 2:00 pm 3:00 pm 2:30 pm

The timing of decisions  Intermediate-term unit commitment problem Turbine 3 Turbine 2 Turbine 1 Notification time 1:15 pm 1:30 2:00 pm 3:00 pm Ramping, but no on/off decisions. Commitment decisions

The timing of decisions  Real-time economic dispatch problem 1pm 2pm 1:05 1:10 1:151:201:251:30

The timing of decisions  Real-time economic dispatch problem 1pm 2pm 1:05 1:10 1:151:201:251:30 Run economic dispatch to perform 5 minute ramping Run AC power flow model

SMART-ISO: Calibration Historical generation mix during 22-28 Jul 2010 Nuclear Steam Comb. cycle+gas Pumped hydro

SMART-ISO: Calibration Simulated generation mix during 22-28 Jul 2010 Steam Nuclear Comb. cycle+gas Pumped hydro

Actual vs. simulated LMPs January April July October

Energy from wind 30 days  Wind power from all PJM wind farms

Energy from wind  Illustration of forecasted wind power and actual »The forecast (black line) is deterministic (at time t, when the forecast was made). The actuals are stochastic. This is our forecast of the wind power at time t’, made at time t. This is the actual energy from wind, showing the deviations from forecast.

Energy from wind  Two types of uncertainty arise in forecasting: »At time t, we have forecasts for different times into the future: The forecast is an imperfect estimate of the actual load at time t’: »As new information arrives, the forecasts themselves change from time t to t+1: This change in the forecast is “stochastic” at time t. The actual load at time t’ is “stochastic” at time t.

Forecasting wind  Rolling 24-hour forecast of PJM wind farms Actual Hour of day 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 Megawatts

Onshore & offshore wind farms  We were given access to data on the wind power generated by onshore wind farms within PJM Proposal: Use onshore data to calibrate a stochastic model of forecasting errors. Then use this model to create a simulated “actual” for offshore.

 Distribution of forecast errors »Uses adjusted spatial correlations to improve fit. Simulating onshore wind Observed Simulated Error in forecasted wind speed

Simulating onshore wind  Cumulative histogram of the # of consecutive time intervals the observed/simulated time series is above the forecasted one (chosen farm only): Time actual is above forecast Observed Simulated

Simulating onshore wind  Cumulative histogram of the # of consecutive time intervals the observed/simulated time series is below the forecasted one (chosen farm only): Time actual is below forecast Observed Simulated

Wind forecast samples 80 70 60 50 40 30 20 10 0

Simulating offshore wind  Offshore wind – Buildout level 5 Forecasted wind

SMART-ISO: Offshore wind study Mid-Atlantic Offshore Wind Integration and Transmission Study (U. Delaware & partners, funded by DOE) 29 offshore sub-blocks in 5 build-out scenarios: »1: 8 GW »2: 28 GW »3: 40 GW »4: 55 GW »5: 78 GW

Modeling wind »Steadier than onshore? Where??? GW

Modeling wind  The power from wind: »The cubic relationship means small changes in speed translate to large changes in power.

Unit commitment under uncertainty Actual wind Hour ahead forecast How forecasting uncertainty causes outages.

SMART-ISO: Offshore wind study

Less steam Uniform increase in gas

SMART-ISO: Offshore wind study  Outage probabilities over 21 scenarios for January, April and October: Percent of samples there is an outage Base PJM reserve Optimized reserves Perfect information

SMART-ISO: Offshore wind study  Outage probabilities over 21 scenarios for July Percent of samples there is an outage

SMART-ISO: Offshore wind study  Ramping reserves, July, 2010 Perfect forecast Imperfect forecast 16 14 12 10 8 6 4 2 0 123 45 Buildout levels Ramping reserves (GW)

SMART-ISO: Offshore wind study  Observations »The requirement of reserves at 20 percent of capacity is only for July, and we believe this over-estimates the reserves needed. »Other months require reserves at 10 percent of capacity, which is still substantial, considering that renewables generate energy at roughly 20 percent of capacity. 10 percent reserves means that 100 MW of wind generation requires 10 MW of spinning reserve. 100 MW of generating capacity translates to around 25 MW of power (on average). So 25 MW of power requires 10 MW of spinning reserve from a fossil plant. »This is for an “as is” network – we are using existing generation technologies and existing planning procedures. »A richer portfolio including demand response, battery storage and more sophisticated planning should help.

The challenge of renewables  What did we learn? »Wind and solar are variable, and uncertain, with very different characteristics. »Back-of-the-envelope analysis (e.g. capacity factor analysis) completely ignores the challenges of dealing with variability and uncertainty. »It is important to design effective policies for dealing with variability, forecasts, and uncertainty. »Reserves to handle uncertainty can be significant, and are easily overlooked. »Renewables are a powerful alternative to reduce our CO2 footprint, but they need to be planned properly to avoid unexpected costs.

Slide 1 The Renewables Challenge: Keeping the lights on while managing variability and uncertainty Princeton University Academic Mini-Reunion October 2,

Similar presentations

Presentation on theme: "Slide 1 The Renewables Challenge: Keeping the lights on while managing variability and uncertainty Princeton University Academic Mini-Reunion October 2,"— Presentation transcript:

Similar presentations

About project

Feedback

Log in

Auth with social network:

Slide 1 The Renewables Challenge: Keeping the lights on while managing variability and uncertainty Princeton University Academic Mini-Reunion October 2,

Similar presentations

Presentation on theme: "Slide 1 The Renewables Challenge: Keeping the lights on while managing variability and uncertainty Princeton University Academic Mini-Reunion October 2,"— Presentation transcript:

Similar presentations

About project

Feedback