From Forecast to Discovery: Applying Business Intelligence to Power Market Simulations Presented at the Energy Central Webinar May 12, 2015 Copyright 2015 Newton Energy Group LLC
Welcome and Introduction Today s Objectives Applying and Capturing the Business Intelligence of Power Market Simulations Today s presenters Richard Tabors, Host President: Tabors Caramanis Rudkevich Visiting Scholar and Co Director, Utility of the Future Project at the MIT Energy Initiative Alex Rudkevich, Ph.D. President: Newton Energy Group John Goldis, Ph.D. Chief Technology Officer: Newton Energy Group 2
Outline Uses of power market forecasting Market simulators as forecast engines: missed opportunities A case study: applying business intelligence to NYISO Market simulations What lies behind the forecast Conclusions Q&A 3
Forecasting THE Forecast is Always Wrong 4
Why? We are always time constrained to finish THE forecast We spend too little effort being able to see the broader implications of our forecasts, i.e., understanding the breadth of knowledge embedded in our simulations We too often lose the intermediate results when we would like to return to the analyses. 5
Typical uses for power market forecasts Valuation of assets (physical or financial contracts) Cash flow projections under various scenarios System planning Assessment of supply, demand and physical flows in the grid, economic and environmental impacts of generation, demand-side and transmission projects. Costbenefit analyses Policy analysis, market design Simulation of the impact of changing regulatory policy, market /operational rules on market performance. Cost-benefit analysis Generation scheduling, trading support Detailed simulations of system operations and economics under multiple scenarios with relatively short-term horizons (hour-ahead to month-ahead) Risk analysis Cash flow projections under multiple scenarios for risk metrics Modeling of variable generation, distributed generation, demand response participation in markets for energy and ancillary services Hourly and sub-hourly simulations of market operations under various inputs and market design scenarios 6
The forecast is always wrong but The value of the forecast is in maintaining an understanding of the fundamental relationships underlying the forecast (i.e. models), formulating assumptions (i.e. inputs) and analyzing their implications (i.e. outputs) A forecast developed by a market simulator (model) simultaneously and consistently predicts a very large number of physical and economic characteristics of the market (outputs) in response to assumed inputs 7
Key Components of Power Market Simulators Inputs Models Algorithms Demand forecasts Generation and Transmission expansion Fuel Prices Emission Allowance Prices Loads, demand response Transmission: existing, new; constraints, contingencies Generation: existing, new; Storage; Variable generation Market rules Maintenance Scheduling; SCUC/SCED; Contingency analyses; Energy and A/S co-optimization; Topology control Outputs (Forecast) Physical: Generation and reserves schedules Power flow Fuel use Emissions Curtailments Financial: Prices Revenues Costs 8
Opportunities Single purpose forecasts typically generate gigabytes of information but ultimately use only a few numbers Most of the information generated by such forecast remain unused requires direct access to the simulator requires specialized training in-house or retaining consultants takes too much time to formulate and execute each request ultimately this just costs too much! Business intelligence (BI) is the set of techniques and tools for the transformation of raw data into meaningful and useful information for business analysis purposes. Use of cloud based Business Intelligence (BI) tools make this information: accessible to a much wider range of users inexpensive to access and navigate 9
A case study: applying business intelligence to NYISO Market simulations
Visit NEG s Power Market Explorer https://pme.negll.com All of the case study results presented below could be obtained via the free demo of the Power Market Explorer 11
The Case Study Forecast NYISO market over a 2-year period: June- 2015 through May-2017 Three Scenarios: Business as usual (BAU) High Load: 90/10 load forecast vs. 50/50 load forecast under BAU High Gas: gradually transition from forward prices as of May 4, 2015 to forecast per 2015 EIA Annual Energy Outlook Modeling technology: NEG s pcloudanalytics cloud based platform Power Systems Optimizer (PSO) simulator by Polaris Power Market Explorer (PME) BI solution by NEG 12
Results Flow Logic BAU case: Prices Generation mix Performance of CCGT and Wind generators Emissions Fuel (natural gas use) Performance of asset portfolios Sensitivity cases: Impact of sensitivity assumptions on the above categories 13
Forecast of Energy Prices Winter spikes in natural gas prices have bigger impact on LBMPs than the difference between summer and winter demand 14
Forecast of Energy Prices (cont d) especially during off-peak hours 15
Forecast of Price Basis (from Zone D) The effect of congestion on price separation within NYISO is more influenced by gas prices than by load 16
Forecast of Price Basis (from Zone D) even more so in off-peak hours. 17
Projected Generation Mix Natural gas plays a dominant role among thermal technology. Coal and oil fired generation will have marginal presence primarily during peak winter and summer periods 18
Comparing Performance of CCGTs by Zone CCGTs generation by month and NYISO Zone. CCGTs are concentrated predominantly in zones A, C, F, J and K 19
Margins earned by CCGTs ($/MW-day) will vary significantly over time and by geography CCGTs in the West (Zone A) will likely outperform CCGTs in NYC (Zone J) and on Long Island (Zone K) 20
Wind Generation (MWh) Significant seasonal variation in wind availability 21
Wind Energy Revenues per MW-day of Installed Capacity On a per MW-day basis, the highest value wind appears in the Capital Zone (Zone F), the lowest value wind in the North Zone (Zone D) 22
Or to track emissions CO2 emissions by zone by month reported in billion lbs Bottom figures show emissions during OnPeak (left) and OffPeak (right) hours 23
or track natural gas use by zone monthly NG use (MMBtu) reported by month by Zone 24
or by pipeline pricing point monthly 25
or by pipeline pricing point daily 26
With BI it is easy to define, analyze and compare portfolios of assets Portfolio 1 (330 MW): 25% of Astoria CC (Zone J) 25% of Athens 2 (Zone F) 100% of Bliss Wind (Zone A) Portfolio 2 (325 MW): 25% of Empire CC (Zone F) 50% of KIAC CC (Zone J) 100% of Northridge Wind (Zone E) 27
High Load Scenario based on the 90/10 Gold Book Forecast modeled as a significant demand increase in peak months (Jan and Jul) and very minor increase in other months It is always good to compare scenarios High Gas Scenario (gradually transitions from the forward to AEO-2015) 28
Higher loads influence LBMPs unevenly in time and geography In months of high load increase (Jan and Jul) LBMPs increase everywhere but not uniformly. Minor load increases sometimes lead to lower prices. The figure shows LBMP increase from the BAU Scenario 29
Impact of High Gas Prices on LBMP As expected, higher gas prices lead to higher LBMPs. However LBMP increases from the BAU Scenario (as shown below) affected by transmission congestion vary by location 30
Incremental generation that serves incremental load is fueled primarily by natural gas 31
Impact of High Gas Prices on Generation Mix 32
Impact of high natural gas prices on natural gas use: increase in realized NG price ($/MMBtu) v. decrease in NG use (MMBtu) 33
Impact of high load on natural gas use: incremental load (MWh) v. incremental NG use (MMBtu) 34
Higher loads benefit CCGT units CCGTs in Zones A, J and K earn additional $23 and $25 per MW-day, respectively For Zones C and F additional earnings are only $13 and $9 per MW-day, respectively 35
Impact of higher load on wind performance Wind in Zones A D will earn additional $11-$13 per MW-day in revenues For wind in Zone F is only $6/MW-day increase 36
Mixed impact of higher gas prices on CCGT performance CCGTs in Zones A, C and K earn additional ~$7.5 per MW-day, in zone K ~$3/MW-day The average effect on CCGTs in Zones F and J will be negative 37
Higher gas prices benefit wind 38
Comparing Portfolios as they respond to higher loads Portfolios 1 and 2 benefit under the High Load scenario by $16.8 and $15.6/MW-Day, respectively Portfolio 1 better responds to summer load increase, Portfolio 2 to winter load increase 39
Impact of High Gas Scenario on Portfolios Portfolios 1 and 2 benefit under High Gas scenario by $3 and $5.5/MW-Day, respectively 40
Copyright 2015 Newton Energy Group LLC Behind the Forecast
Data Outputs The Making of a Forecast Expertise Business Intelligence 42
Data and Expertise Data RTO data and studies EIA EPA NREL EvoMarkets FERC NERC Commercial data vendors (e.g. SNL) Expertise Data Power systems planning and operation Market rules Fuel and emission markets Modeling IT Simulator Inputs, models and settings 43
Simulator A great Simulator closely emulates market clearing models used by the ISOs and is able to reproduce the physical reality resulting from market clearings Key Requirements for Modern Simulators Mixed integer programming (MIP) Capture increased system and market complexity Results shown on previous slides were generated with Power System Optimizer (PSO), a state of the art simulator built on AIMMS and powered by the Gurobi solver. 44
Why Mixed Integer Programming? ISOs are using MIP algorithms in their unit-commitment and dispatch processes Previous methods relied on approximations to the physical market problem yields suboptimal results Previous software and hardware limitations prevented MIP algorithms from finding good solutions within required time limits From 1991-475,000x algorithmic speedup - 2,000x machine improvement Total Improvement - ~10 9 x speedup What used to take a million years can now be done in a few hours MIP is theoretically capable of finding an optimal solution and practically performs better Lower production costs Increased modeling flexibility and addressing system and market complexity Increases support for future market design changes PJM and SPP estimated $100 million in annual savings due to MIP implementation NYISO estimates that ability to properly model combined cycles with a MIP implementation could impact LBMPs by 11% 45
System and Market Complexity There are many complex resources in the market Combined cycle model CCs and some other generators can operate in several configurations Dispatchable demand response direct load control for reliability or economic reasons Storage / storage + renewables facilities that combine wind or solar with on-site storage Flexible and Dynamic Reserves Customizable reserve products differentiable by up vs. down requirements Changing transmission topology Need to represent resources at a sub-hourly scale Evolving market products and rules As the market evolves and rules change, the simulator should adapt and represent these changes explicitly 46
Business Intelligence (BI) Business Intelligence is a set of techniques and tools for the transformation of raw data into meaningful and useful information that is easy to access. Simulators generate 10s to 100s of GBs of raw data per simulation (scenario) For proper comparative analysis involving large number of scenarios may have TBs of data This raw data is made meaningful when it can be summarized and grouped into useful categories Generation by technology type Emissions by fuel Prices by control area Market heat rates Revenues and costs by portfolio Congestion patterns by zone Analysts already use Excel for many business analytic needs - PivotTables and PivotCharts are intuitive and established tools for organizing information Self-service BI Easy to create templates and compare results across scenarios The power of the Cloud Allows data to be centrally and securely stored Can deliver BI solutions directly to analysts with minimal additional software 47
Power Market Explorer as a Self-Service BI To continue exploring the NYISO results, register for free at: https://pme.negll.com You will receive an Excel workbook with some pre-generated templates and can get started creating your own reports and comparing results between the business-as-usual, high load and high gas scenarios for the June-2015 through May-2017 time period. 48
Summary and Conclusions We can be far smarter in our use of Power Market Simulators to generate and thereby be able to use more Business Intelligence Focus on additional variables that allow us to better understand the economic and physical systems not just a single value Retain and analyze (and compare) multiple cases Use state-of-the-art technologies to make data and analyses into Business Intelligence 49
Q&A 50
Contact Us Richard Tabors 617-871-6913 rtabors@tcr-us.com Alex Rudkevich John Goldis 617-340-9810 617-340-9815 arudkevich@negll.com jgold@negll.com 75 Park Plaza, 4 th Floor Boston, MA 02116 www.newton-energy.com 51