Metric and Dispatch Approach for Deliverable Ramping Flexibility Anupam Thatte, Midcontinent ISO Dr. Le Xie, Texas A&M University College Station, TX Nov 2016 1 A. A. Thatte and L. Xie, "A Metric and Market Construct of Inter-Temporal Flexibility in Time-Coupled Economic Dispatch," in IEEE Transactions on Power Systems, vol. 31, no. 5, pp. 3437-3446, Sep. 2016.

Outline •Motivation •Probabilistic Metric for Ramping Flexibility •Proposed Two step Dispatch Approach •Numerical Case Study •Summary 2

• Increased uncertainty and variability due to large scale deployment of renewables • Challenge for power system operators to maintain power balance despite ramp events • Ramp shortages can lead to price spikes, and require expensive out-of-market actions by operators • Thus, ramp flexibility is a key issue for system operators Motivation 3

Flexibility Requirement in Dispatch •Need adequate ramp to cover variability and uncertainty in dispatch horizon 4 Uncertainty in Net Load (based on 99.7% CI i.e., ± 3σ) Time Net Load t-1 t t+1 t+2 t+3 Point Forecast Uncertainty Real-time economic dispatch (5 min intervals) Fig. Source: MISO

Probabilistic Metric for Ramping Flexibility •We propose Lack of Ramp Probability (LORP) as a metric for flexibility in economic dispatch •Motivated by the Loss of Load Probability (LOLP) metric •Measures the probability of ramp shortage in real- time economic dispatch Can be used to compare various real-time dispatch models (conventional, look-ahead, robust, etc.) Impact of other resources (such as Demand Response, Energy Storage) on system ramp capability can be evaluated using this metric 5

Lack of Ramp Probability (LORP) • Using forecast data we can obtain pdf of net load in next interval (red curve) • Solid blue line indicates the system dispatch level in current interval • Dotted blue line is the level to which system can ramp up (in one interval) • Green area sums to probability of ramp shortage occurring • Similarly for Ramp Down case 6 Lack of ramp Probability Σ Gen Outputs [t] Σ Ramp [t] Pdf of Net Load Point Forecast of Net Load [t+1] MW

Flexibility in multi-zonal systems •System wide LORP does not provide full picture •Ramp shortages can occur due to both insufficient generator output as well as transmission constraints 7 Source: CAISO, Report on Market Issues and Performance, Feb 2014

Market construct for multi-zonal systems • Use bus aggregation to obtain reduced system [Shi and Tylavsky, 2012] • Each zone aggregated as one node (all generators and loads connected to one node) 8 Step 1: Robust dispatch on aggregated system Inter-zonal Tie-Lines Wind Wind Wind Coal Coal Gas Gas Gas Gas Gas Nuclear Nuclear Coal Bus 1 Bus 2 Bus 7 Bus 5 Bus 8 Bus 4 Bus 3 Bus 9 Bus 10 Bus 6 Bus 24 Bus 11 Bus 12 Bus 13 Bus 14 Bus 15 Bus 16 Bus 19 Bus 20 Bus 21 Bus 22 Bus 18 Bus 17 Bus 23 Oil Oil North East South West

North East South West Inter-zonal Tie-Lines Market construct for multi-zonal systems • Consider uncertainty set for zonal net load in future time interval (i.e. t+1) • Consider inter-zonal tie-line flow limits 9 Step 1: Robust dispatch on aggregated system

Market construct for multi-zonal systems • Maintain Tie-line flows from Step 1 (within some tolerance) • Ensure ramp capability for all zones • Under Gaussian uncertainty assumption, the uncertainty set internalizes the desirable LORP requirement. • We solve for the current time interval (t = 1) and future time intervals (t = 2, 3, . . . , T ), which are considered advisory. 10 Step 2: Deterministic dispatch on full system Step 1 Step 2 Information 1. Tie-Line flows 2. Zonal ramping capability Robust Dispatch Deterministic Dispatch

Case Study 3 zone 73 bus IEEE RTS system • Proposed approach has higher dispatch costs (without considering cost of out of market actions, which can be very expensive) • Lower probability of shortage events, both system-wide and zonal 11 Mean LORP for given day Dispatch Approach Conventional Look- ahead Two-step Robust Dispatch Costs (Mn $) 41.77 41.85 50.82 Mean LORPup 0.0873 0.0821 0.0013 Mean Price ($/MWh) 70.39 69.53 46.45 Max Price ($/MWh) 664.05 676.85 464.72 Matlab Time (s) 1.96 2.7 Steps: (1) 67.38 (2) 0.7

Summary •Lack of Ramp Probability (LORP): A metric for system operators to manage ramp flexibility • Rigorous and intuitive • (1) System-wide, and (2) Zonal •Two-step robust dispatch is formulated • Robust dispatch on average gives lower number of shortage events • Robust dispatch gives lower LORP • Two-step approach ensures deliverability of ramp capability across tie-lines 12

Key References • D. Bertsimas, E. Litvinov, X. A. Sun, J. Zhao and T. Zheng, "Adaptive Robust Optimization for the Security Constrained Unit Commitment Problem," IEEE Trans. Power Syst., vol. 28, no. 1, pp. 52-63, Feb. 2013. • A. A. Thatte, X. A. Sun, and L. Xie, “Robust optimization based economic dispatch for managing system ramp requirement,” in Proc. 47th Hawaii Intl. Conf. on System Sciences, Hawaii, HI, Jan. 6-9, 2014, pp. 2344–2352. • D. Shi and D. J. Tylavsky, “An improved bus aggregation technique for generating network equivalents,” in Proc. IEEE PES General Meeting, San Diego, CA, Jul. 22-26, 2012, pp. 1– 8. 13

Thank you Anupam Thatte Engineer, Market Evaluation & Design Department, Midcontinent ISO athatte [at] misoenergy [dot] org 14

Two step real-time dispatch 15 Proposed approach considers • uncertainty and variability in zonal net load • inter-zonal tie-line flow constraints • the deliverability of procured ramp capability between operating zones

Implementation •For each interval generators will submit their supply bids for time t and t+1 •ISO will run the two-step optimization model •At each time interval • Dispatch solution for t will be binding • Dispatch solution for t+1 will be advisory •LMPs will be obtained from step 2 (deterministic dispatch) •Note: Uncertainty set selection in step 1 is done by ISO 16

Zonal Ramping Flexibility 17 Source: Potomac Economics, ERCOT State of Market Report 2013 • In ERCOT most of wind generation is in West and South, whereas load is in North and Houston • System-wide approach may not be enough to guarantee flexibility

Inter-zonal transmission constraints • We can treat the inter-zonal tie-lines as virtual generators/loads 18 Zone Virtual gen Virtual load • Ramping metric that observes tie-line flow constraints • Ensure deliverability of ramping in multi-zonal systems

19

Step 1: Robust Dispatch on aggregated system 20 Minimize costs Power Balance Worst case net load Generator limits Ramping limits Tie-Line limits Tie-Line limits Uncertainty sets

Step 2: Deterministic Dispatch 21 Minimize costs Power Balance Generator limits Ramping limits Zonal ramping capability Observe Step 1 Tie-Line limits Transmission limits

Increasing Penetration of Renewables • Increase in uncertainty and variability impacting the power system balancing 22 58 79 105 160 298 382 852 1922 3369 4776 6201 0 1000 2000 3000 4000 5000 6000 7000 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 MW Year Solar PV 372 2396 2454 5237 8425 9918 5112 6649 13089 1102 4772 0 2000 4000 6000 8000 10000 12000 14000 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 MW Year Wind Source: US DOE, 2014 Renewable Energy Data Book

• The short-term wind forecast error makes it challenging to ensure adequate ramp from conventional generators Challenge of Ramp Shortages 23 Source: CAISO, Report on Market Issues and Performance, Feb 2014

Lack of Ramp Probability (LORP) • Using forecast data we can obtain pdf of net load in next interval (red curve) • Solid blue line indicates the system dispatch level in current interval • Dotted blue line is the level to which system can ramp up (in one interval) • Green area sums to probability of ramp shortage occurring • Similarly for Ramp Down case 24 System Current Output at time t System Output + Ramp at time t+1

Two step robust dispatch framework • Step 1 = robust dispatch on aggregated zonal model • Step 2 = deterministic dispatch on full model 25 Zone Tie line flow Tie line flow • Convert system to aggregated model • All generators and loads in a zone aggregated at one bus • Proposed approach considers tie-line constraints • Ramp requirement for zone is considered North East South West Inter-zonal Tie-Lines Step 1 Step 2 Information 1. Tie-Line flows 2. Zonal ramping capability A. Thatte and L. Xie, “A Metric and Market Construct of Inter-temporal Flexibility in Time-coupled Economic Dispatch,” IEEE Trans. Power Systems, 2015 (accepted)