MetrixIDR Retail
Return to Energy Forecasting
Energy retailers need timely and accurate forecasts to schedule their loads in day-ahead
and balancing markets. Energy traders need accurate forecasts to manage their risk.
MetrixIDR Retail generates sub-hourly, hourly, or daily forecasts for portfolios of electric
and gas retail customers or lists of delivery points, where the portfolio can be
changed on a daily basis.
For profiled customers, the system uses the profiling method employed in a particular
market region. For individual customers or delivery points, the system provides
a set of template forecasting models that are assigned based on the properties of
the load. This allows simple models to be used for calendar-driven loads, while
more complex models are used for loads driven by weather, price, or other factors.
For the largest customers or delivery points, customised forecasting models for
each customer is utilised.
Included in MetrixIDR Retail is a built-in scoring algorithm to help identify the best
model type for each case. The scoring statistics generated quantify load variability
and the importance of calendar effects, seasonal effects, daily weather sensitivity,
and price sensitivity. Model forecasts for individual customers, groups of customers,
or profile segments are mapped to schedules for aggregation purposes. Typically
one schedule will be defined for each transmission zone or control area in the portfolio.
MetrixIDR Retail is implemented in Itron's
Retail Forecasting Systems and
Delivery Point Forecasting Systems. Employing MetrixND® as the forecast engine, Itron's systems
allows the user to assign preconfigured modelling templates to each forecasting
task, or to develop a customised model for a specific forecast element.
Itron understands the complexities of evolving markets and the importance of accurate
forecasts in these markets - take advantage of our knowledge and solutions.
System Features
Customised Configurations
Configured for retail forecasting with profile models, individual customer models,
or both. It can be configured to forecast wholesale loads for individual delivery
points or groups of delivery points. The system can also be scheduled to update
as needed during the day.
Modelling Flexibility
The user manages lists of customers or delivery points. Customers may be modelled
individually, or the customer loads can be aggregated into a meter group, allowing
forecast models to be applied to the group as a whole. A wide range of modelling
methods is available and there is no limit to the types of explanatory variables
that can be used.
Built-in Scoring Algorithm
A built-in scoring algorithm is included to help identify the best model type for
each case. The scoring statistics generated quantify load variability and the importance
of calendar effects, daily weather sensitivity, and price sensitivity. As an option,
rules can be implemented to automatically assign methods to each forecast element.
Rapid Model Estimation
Once a method is selected either manually or by automatic assignment, the model
is populated with data from the meter or profile database tables, model weather
data for the appropriate weather zone, the forecast model coefficients are estimated,
and the estimated model is saved for use in forecast generation. These steps take
a few seconds for each model that needs to be estimated. Estimated models can be
exported for detailed analysis and development of refinements to the specification.
Rapid Forecast Execution
Estimated models are used to generate forecasts on a scheduled basis or on demand.
Forecast execution takes a few seconds per modelled item. If recent data are available
for calibration, forecasts are adjusted based on this recent history, using the
dynamic learning algorithm.
Flexible Forecast Schedules
Model forecasts for individual customers, groups of customers, or profile segments
are mapped to schedules for aggregation purposes. Typically one schedule will be
defined for each transmission zone or control area in the portfolio.
Weather Forecasts
Import via the Internet, weather forecasts from one or more weather stations and
one or more weather service providers. All weather concepts are supported including
temperature, humidity, dew point, cloud cover, wind speed, and precipitation.
Prior Hour Loads
The system can be configured to use real-time meter reads to drive the forecast.
Error Trapping
Built-in validation routines flag errors in weather and load data.
On-line Review
A variety of on-line Quick Edit tools are available for refining the model forecasts.
Publishing
Forecast results and corresponding weather data can be published by the forecast
operator to a variety of file formats including custom formats to support downstream
applications.
Integrates with Existing Databases
Tailored to work with your existing databases, including Access®, MV-90, MV-Star,
ORACLE®, SQL Server®, and SAP®.
Call us at +31 20 60 65 225 or send an email to forecasting@itron.com for more information or to schedule
a live software demonstration.
More Information
