SMART GRID FORECASTING

Transcription

1 SMART GRID FORECASTING AND FINANCIAL ANALYTICS Itron Forecasting Brown Bag December 11, 2012

2 PLEASE REMEMBER» Phones are Muted: In order to help this session run smoothly, your phones are muted.» Full Screen Mode: To make the presentation portion of the screen larger, press the expand button on the toolbar. Press it again to return to regular window.» Feedback and Questions: If you want to give feedback to the presenter during the meeting or if you have a question, please type your question in the Q&A box. We will address it as soon as we can.

3 BROWN BAG SEMINARS 2012» Using Conditional o Heteroskedastic edast Variance a Models in Load Research Sample Design - March 6» Commercial Sales Modeling and Forecasting June 19» Forecast Accuracy vs. Forecast Stability Sept 18» Smart Grid Forecasting Dec 11, » Solar and Wind Generation Modeling and Forecasting - March 5» 2013 Forecast Accuracy Benchmarking Survey and Energy - June 4» Update on Weather Normalization Trends and Issues - Sept 17» 2014 Outlook - Dec 10, 2013 All at noon, Pacific Time All are recorded and available for review after the session.

4 Smart Grid Forecasting and Analytics» AMI and Smart Grid deployments and technologies Pose new forecasting challenges, but also Enable new forecasting and analytical possibilities» The presentation covers three specific areas: Operational forecasting for the distribution grid Analytics for financial closing Analytics for budget tracking and weather adjustment» The benefits Enable operational efficiencies Achieve financial clarity 4

5 1. Operational Smart Grid Forecasting» System that generates short-term operational forecasts Short term (today, tomorrow, next few days) Interval level (hourly or sub hourly) Updated frequently (e.g., every 15 minutes) At the substation and/or feeder level» Accounts for emerging technologies Embedded solar and other distributed generation (DG) Load reductions available through Demand Response (DR) Load impacts from electric vehicle charging Other important embedded technologies» Interacts with and informs other automated systems Distribution management (DMS) and automation (DA) systems Demand response management systems (DRMS) 5

6 What makes Operational Forecasting Work» There is a strong, stable relationship between weather and load» Fundamental drivers do not vary from day to day Economy Customer base End-use equipment» Calendar explains the base» Weather explains the lift Temperature is key Humidity matters Wind matters Clouds matter Three Years of Daily Data Each Point is One Day Weekdays Sundays Saturdays Holidays 6

7 Operational System Forecasting Today Generation dispatch decisions are made based on SCADA data from generation stations ti and transmission i system tie points Load ISO, TSO, RTO, Utility Ops -- Day ahead markets -- Hour ahead markets 7

8 Tomorrow -- Moving to the Grid Level ISO, TSO, RTO Component» Requires grid measurement data at the substation or below» Accounts for emerging technologies and programs enabled by Smart Grid initiatives 8

9 Operational Grid Forecasting = Grid Loads Substation Loads Feeder Loads Traditional Loads Embedded Solar and other DG Electric Vehicle Charging Demand Response Impacts Inputs: Weather Calendar Data Economics End Use Data Inputs: Cloud Cover Installed Capacity T t AMI Output Data E i Temperature, Inputs: Rate Structures Saturation Charging voltage Charging location AMI Charging Data Inputs: DR Programs Event Days Rates Installed Technology AMI Load Data 9

10 ELECTRIC VEHICLE FORECASTING» Electric Vehicles (EV) and Plug in Hybrids (PHEV) are a new electric end use» With 30 miles per day driving, an EV is like an electric water heater it adds about 4 MWh/year to a residential load. MW 4,500 4,000 3,500 3,000 MW2,500 2,000 1,500 1, Midnight Super Off peak The main questions are when and where will charging occur. Off peak On peak Off peak Noon Midnight» Rates can strongly impact the charging profile.» Separate metering supports smart charging strategies J. Wynne and Dr. L. Pratson, Impact of Plug-In Hybrid Electric Vehicles on California s Electricity Grid, May

11 Distributed Generation View 7 Day View 1 Day View DG View DR View Green Tree Substation Current Time: Mon, Jul 23, 06:15 Generation in MW Cloud Cover Index Hourly Temperature 11

12 Demand Response View 7 Day View 1 Day View DG View DR View Green Tree Substation Current Time: Mon, Jul 23, 06:15 DR Available in MW DR Available in MW C&I Program (MW) Res Program (MW) 12

13 SG Forecasting Results What you get 7 Day View 1 Day View DG View DR View Green Tree Substation Current Time: Mon, Jul 23, 06:15 Gross Usage (MW) Net of DG (MW) DR Available (MW) Hourly Temperature 13

14 1 Day View 7 Day View 1 Day View DG View DR View Green Tree Substation Mon, July 23 Current Time: Mon, Jul 23, 06:15 DG (MW) Gross Usage (MW) Net of DG DR Available DR (MW) Gross Net DG Net DR DG DR Energy (MWh) 3,605 3,515 3,405 Energy (MWh) Peak (MW) Peak (MW) PeakTime 16:15 17:00 17:30 14

15 Smart Grid Forecasting System Interactions = Grid Load Forecasts Traditional Loads Embedded Solar and other Distributed Generation Electric Vehicle Charging Load Load Data Forecasts Program Data Demand Response Availability DR Impacts Distribution Management and Automation Systems (DMS & DA) End-use Saturation, ti Energy Efficiency DG Output Metering, AMI Data Analysis EV Metering, Rates, AMI Data Analysis Demand Response Management System DRMS), AMI Data Analysis 15

16 WHAT ITRON IS DOING» The Itron Automated Forecasting System: Widely used at the ISO, TSO, Utility level. Scalable and meets the functional and performance requirements for operational forecasting at the grid level. Status: Working with our customers to understand emerging needs, possible applications, and benefits of forecasting at the grid level.» Working with utility customers on demand response modeling and forecasting of DR availability.» Working with ISOs on improved algorithms for solar and wind forecasting.» Working on electric vehicle issues for long-run forecasting purposes.» Working on a platform to support AMI analytics that can inform the forecasting process. 16

17 Part 2: AMI Data and Financial Closing» Billing is by cycle» Financial reporting is by calendar month» Each month analysts must estimate: Calendar month sales and revenue Unbilled sales and revenue» This can be a pain point depending on the methods used» AMI data allows replacement of estimates with direct measurement» Partial AMI data can improve estimation process during deployment 17

18 The role of billing cycles» Most utilities use 20 or 21 billing cycles» Some use 40 bimonthly cycles» Not always a clean precise process with manual reads or AMR 18

19 Why AMI data clarifies the process» AMI data provides day-by-day usage for each customer» AMI data is up to date for closing calculations» AMI is the source of usage data for computing bills» With AMI data Calendar month energy can be calculated directly Unbilled energy can be calculated l directly» Complicated estimation processes can be replaced by direct calculations 19

20 Why AMI data clarifies the Process» Complicated estimation processes can be replaced by direct calculations The Close Calendar Month Energy A B C Billed Energy Unbilled Energy 20

21 Using AMI data during deployment» There are several estimation methods using aggregated or detailed billing data.» All methods use profiles to allocate and extend bills Calendar month/billing month ratio ((B+C)/(A+B)) Unbilled/calendar month ratio (C/(B+C)) C))» Partial AMI data can be used to generate profile data and ratios during deployment» We have looked at this using data from the 30% deployment point and concluded that partial AMI data compares favorably with load research data and can be used to improve the closing process. 21

22 Part 3: Budget Tracking & Weather Response» Looking back at each month, weather effects are quantified to: Analyze variances from budget Track weather adjusted sales against budget» Weather effects are often modeled with monthly sales» Use of daily AMI data can strengthen these models» The result is: Stronger, more robust models More precise estimation of weather effects Improved clarity about the business 22

23 Understanding Weather Response AMI data is very powerful for understanding weather relationships. There are 365 points per year, a strong advantage over monthly data. AMI usage on adaycan be matched cleanly with weather for that day. One Year of Daily Data Each Point is One Day Weekdays Sundays Saturdays Holidays 23

24 Why this is Important Most analysis models are built with monthly data. Monthly sales data are from staggered billing cycles. Monthly data have less leverage (in X and Y directions). Daily data show how weather works. AMI Data for one Year. Each point is one day. Billing Data for one Year. Each point is one month. 24

25 AMI Data Supports Richer Models» AMI data shows us when weather starts to matter» AMI identifies low, medium, and high powered degrees» AMI data shows seasonal differences in weather response that are not visible with monthly data.» AMI data identifies weekend effects» AMI data identifies the strength of holiday effects» Understanding these effects creates additional clarity 25

26 AMI Data Supports New Analyses» Models of daily AMI data are useful» Class sales can be adjusted d for weather each day» Class sales can be tracked against budget each day Adjusted Actual ( ) is close to budget ( ). This is good. 26

27 So What?» You put it all together and new possibilities emerge» Improved financial analytics don t increase revenue» They won t decrease cost» But when management talks about results There will be no confusion about how the business sits This will relieve a major pain point in many companies Upper management likes this 27

28 The Bottom Line» AMI data provides improved clarity about customer usage patterns and this supports new and stronger financial analytics. System Load System Peak Day Losses & UFE Delivered Energy Sat Sun Mon Tue Wed Thu Fri 28

29 What this means for Forecasting» Estimation processes for calendar month energy and unbilled energy are replaced by direct calculations.» Budget forecasting needs remain the same. But models will be estimated using daily or calendar month data» Weather normalization processes will improve. These processes will be applied at the daily level, providing improved clarity and visibility. Daily tracking will insure that there are no surprises at month end.» Monthly variance analysis processes will improve. Variance calculations by class will be based on actual calendar month usage instead of estimated calendar month usage. 29

30 What Itron is Doing» Itron Analytics Platform to support analysis of high volume AMI and Grid data Transformer load management Voltage and power quality Revenue protection Dynamic load profiling Financial analytics» Working with utility forecasting groups on Working to quantify operational grid forecasting benefits Using AMI data during deployment Improving forecasting processes Improving financial closing processes 30

31 QUESTIONS? 2013 HANDS-ON WORKSHOPS Press *6 to ask a question» Forecasting 101 April Washington, DC» One-Day Modeling Workshop Commercial Forecast Modeling April 16 - Las Vegas, NV» One-Day Modeling Workshop - Introduction to SAE April 16 - Las Vegas, NV» European Energy Forecasting Workshop - May Brussels, Belgium» Advanced European Energy Forecasting Workshop - May Brussels, Belgium» Fundamentals of MetrixND - May 1-2- Boston, MA» Fundamentals of Short-term and Hourly Forecasting September San Diego, CA» Fundamentals of Sales and Demand Forecasting - October Boston, MA» Forecasting 101 October San Diego, CA OTHER FORECASTING MEETINGS» 7th Annual ISO/RTO Forecasting Summit April Norristown, PA» Energy Forecasting Meeting and Training April Las Vegas, NV - One-Day Modeling Workshop Commercial Forecast Modeling April 16 - One-Day Modeling Workshop - Introduction to SAE April 16-11th Annual Energy Forecasting Meeting April » European Forecasting User Meeting - May Barcelona, Spain» Itron Utility Week October 7-8 Orlando, FL» Australian Forecasting User Meeting - Date TBD - Sydney, Australia For more information and registration: Contact us at: , or forecasting@itron.com