Data Integration and Automation with AURORAxmp
Matthew Jennings, Senior Research Analyst at global energy consulting and research company Wood Mackenzie, presents techniques and strategies for getting results with AURORAxmp. In this discussion, he describes setting up runs, working with outputs, and creating automated reports.
Using such conventional business tools as MS Excel, WinZip and Visual Basic for Applications on standardized files, Jennings describes how to set up runs in an organized storage environment. He then shows how scripts can import AURORAxmp's outputs into a single Access or SQL database, producing aggregate reports.
Once data is in a single output database, Jennings shows how VBA macros, managed by a standard Access “control” file, can export Excel-based reports. These macros link the input and output databases, generating Excel files and cleaning up the data.
Once configured, this entire process occurs seamlessly in the background. Final reports are consistently generated, giving you more time for analysis instead of cutting and pasting. Example final reports conclude the presentation.
Day-Ahead Forecasting with AURORAxmp
Bill Carr, Senior Director, Transmission Analysis at Dynegy, uses AURORAxmp for short term, day-ahead forecasting.
In this presentation, he describes how AURORAxmp is front-loaded with transmission data from OASIS sites, with load, outage, fuel pricing and weather data from third-party vendors, and with load and fuel pricing and weather data from internal resources. The model is updated manually and automatically via linked data tables.
The updated model runs using AURORAxmp’s output templates. AURORAxmp gathers day-ahead and realtime numbers from the major trading hubs, and Carr reviews the numbers to make sure they make sense for the modeled time period.
Using this method, Carr reports, the average time to create a day-ahead forecast breaks down to about 2 hours for data assembly, 30 minutes to update pre-market data, and 30 minutes to run the model and review the results. Dynegy's goal has been to provide a +/- $2 price range for each trading hub. To date, they have managed a 5-7% error rate.
Forward Curves and Price Forecasts
At our 2003 conference, FPL Energy (now NextEra Energy Resources) Director of Market Analysis Bobby Adjemian described the difficulties of predicting future spot power prices.
Predicting market conditions and pricing means using either market-driven forward price curves or model-based simulations. Neither approach is ideal. Forward curve predictions are less reliable on the high side, during peak months, due to suppliers' reactions to actual current market conditions and consumers' aversion towards future risk. Forecast models' solution logic, designed primarily to meet energy demand, are biased to predictions that are lower-than-actual for on-peak times and higher for off-peak.
Comparing the two approaches, FPL found the simulation model a more reliable predictor of future spot prices, when adjusted appropriately to account for the models' biases. However, spot price prediction remained less than adequate: forward curve-based analyses failed due to illiquidity in the power market, and forecasts based on modeling fundamentals had difficulty with bidding strategies and trading dynamics.
Nonetheless, Adjemian found the simulation model approach, when results are interpreted with the approach's inherent biases in mind, to be the most reliable predictor of spot prices.
Using AURORAxmp to Evaluate Transmission Expansion
When expanding transmission capacity, several stakeholders and regulatory bodies must be persuaded to support the initiative. AURORAxmp can be used to create the different analyses these stakeholders demand.
In this discussion, EPIS presenters use AURORAxmp to model the benefits of a hypothetical 500 kV transmission line between Wyoming and Southern Nevada. The valuation model is limited to ratepayer benefits as part of the CPCN application, with a focus on exporting low-cost power and increasing grid efficiency. The model examines the change in cost-to-load to all U.S. ratepayers in the WECC and explores the hypothetical project's value over various futures using scenarios and stochastic analysis.
AURORAxmp demonstrates following zonal and network modeling capabilities:
- Network model and power flow optimization
- Detailed generator representation for commitment and dispatch
- Ability to treat key value drivers as random variables
- Advanced data management and scenario handling
- Quick simulation turnaround
AURORAxmp’s ability to efficiently handle multiple input and output data scenarios can also improve analysis even at an hourly timeframe, given enough processor power and disk space.
This hypothetical focused on one aspect of the transmission project. AURORAxmp can also help analyze:
- Transmission expansion
- Long term capacity expansion
- RPS compliance cost benefits
- Line loss modeling
The demonstrated analysis can also be enriched with the following enhanced stochastic inputs:
- Including all key drivers
- Running a number of iterations determined by statistical testing
- Including short-term and long-term (two-factor) processes
Using AURORAxmp to Model Cogeneration Units
Regulatory & Cogeneration Services, Inc. uses AURORAxmp to model and value cogeneration plants. The analysis evaluates a hypothetical plant in three dispatch configurations: a non-dispatchable combustion turbine, the same configuration with additional dispatchable combustion capacity, and with a dispatchable steam turbine.
AURORAxmp determines revenues, value and dispatch cost ($/MWh) for all three configurations, based upon generation levels, fuel cost and consumption rate.
What Does Wind Really Cost?
Ray Bliven, Power Rates Manager for the Bonneville Power Administration, describes the practical pitfalls of developing wind resources, how to model wind generation with AURORAxmp, and discusses issues surrounding wind generation.
Bliven uses AURORAxmp to model and value a proposed 83-tower wind farm in Eastern Washington. To model the wind resource, AURORAxmp accepts several inputs, including the site's hourly generation or the wind speed and the projected tower heights. From these, AURORAxmp produces the power curve (a cube function of the wind speed) and normalizes the average output. With the initial model built, a representative week for each month is selected, and AURORAxmp converts each hour’s generation into a forced outage factor for each of the 168 hours per week, using a monthly time series vector to refer to the appropriate weekly time series vector.
With this initial model set up, Bliven shows additional analyses. First, using AURORAxmp, he demonstrates the effect of geographic concentration of wind resources on the northwest power market versus geographically dispersed wind assets, in terms of production and pricing.
Bliven then tests how wind generation changes the amount of GHGs produced. (If additional wind generation shifts system operations from base load plants to cycling plants, wind power may paradoxically increase emissions, owing to inefficient dispatching of combustion resources to cover down times.) Blevin uses AURORAxmp’s closed-system dispatch feature to examine two portfolios: one with a 100 MW of wind generation and a 100 MW single combustion turbine, the other with a 100 MW combined cycle combustion turbine. AURORAxmp's analysis shows an increase in fossil fuel usage and GHG production for the non-wind CCCT portfolio, but a significant decrease in NOx and SO2 production.
Bliven then uses AURORAxmp to compare an open system with 4,135 MW of wind generation replaced by combined cycle combustion turbines, finding a smaller increase in fuel usage and GHG production in the non-wind system, along with slight increases in NOx and SO2 production.
Using AURORAxmp, Bliven's analyses indicate that:
- Wind generation in the WECC is substantial enough to affect pricing and warrant hourly modeling to predict market clearing prices
- Geographic concentration of wind generation is more expensive to MCPs (market clearing prices) than dispersed wind generation.
- Emission savings from wind generation are minimized if SCCT (single cycle combustion turbine) operations are increased to regulate wind output and meet load requirements.