Loading...
Case Study: Electric Power2018-11-22T08:12:37+00:00

ADMIE is the Greek Independent Power Transmission Operator (TSO) and has the role of transmitting electrical power at a national level via the power grid. As such ADMIE performs the duties of system operation maintenance and development, in order to ensure Greece’s electricity supply in a safe, efficient and reliable manner, while promoting the development of competition in the Greek electricity market and guaranteeing the non-discriminatory treatment of System users.

Among ADMIE’s responsibilities is the production of optimal power generation schedules on a daily basis (Day Ahead Scheduling and Dispatch Schedules) in the context of the Greek energy market. These schedules are produced by an optimization engine that takes into account information that typically includes load forecasts, renewable sources forecasts, bid quantities and prices submitted by the market participants, interchange schedules and technical information of generation units.

The requirement for a weather forecast

The Operation Planning Section (OPS) of the Department of System Operation and Control is responsible for providing forecasting information to the market platform, which produces the optimal schedules.

Among the forecast data provided by OPS is the system load forecast and the forecast of power produced by renewable sources. System load forecast and forecast of power produced by solar renewable sources (for brevity we shall refer to both as load forecast models) are produced by ADMIE utilizing proprietary software tools owned by ADMIE.

Once the market is cleared, the system marginal prices are produced. Inaccurate forecasts may lead to sub-optimal schedules and it turns out that this has a significant impact in the economic cost of the operation schedules. More specifically, under-forecasts may lead to purchase of expensive services to deal with peaks and over-forecasts may lead to unnecessary capacity being committed.

In effect, load forecasting is fundamental in utility operation and increasing penetration of renewable sources has caused a significant change in the resource mix making the use of accurate forecasts necessary. These changes have constituted load forecasting to be a dynamic process that should continuously be improved.

Weather conditions are among the predominant factors that affect electrical power consumption, and are therefore used as predictors in short-term load forecasting (the weather variables are used differently in long term load forecasting, since many important weather variables are hard to be predicted beyond two weeks). Large penetration of solar generation at the distribution network has created new challenges for load forecasting, as the negative load from distributed energy sources has to be taken into account. This has created the need to incorporate additional weather factors such as light intensity or cloud coverage to accommodate for these effects and divert away from the traditional load forecasting models that used only temperatures or humidity.

Thus weather forecasts are directly linked with energy load forecasts and the accuracy of the first affects the later. It is therefore important for the load forecast to be provided with reliable weather forecasts, as detailed as possible (e.g. hourly) updated frequently throughout the day to further increase their accuracy.

The requirement for a “new” weather forecast

There are four factors that drive the upgrade and reengineering of the requirement according to a new weather forecast engine:

  1. The large penetration of solar energy sources both at the system and the distribution. The light intensity measures that are currently available for four areas in Greece, used by the solar energy prediction model are not sufficient to capture the dispersed nature of solar production.
  2. The need for a realistic, accurate and integrated consideration in the forecast of all renewable sources, so that, together with the production by the conventional units the actual load demand can be forecasted.
  3. Upcoming amendments in the Greek Grid and Exchange Code will require renewable sources to participate in the electricity market. This has significant effects in the load forecast as the power produced by these sources will no longer be treated as negative load.
  4. ADMIE recently implemented a new EMS, which provides a Load Forecast module, the specifications of which required weather variables that were, at the time, not available.

The aforementioned factors imposed ADMIE to proceed with a tender in order to purchase a weather forecast service that would provide a reliable and unified weather data source to cope with the new requirements.

Ex Machina – IBM Partnership

IBM partnered with Ex Machina in order to provide the best possible solution available as per ADMIE’s requirements. For this solution EXM contributed its expertise and experience in the weather forecasting data analytics, combined with the deep knowledge of PaaS related cloud infrastructure.

The provided solution is Weather Ex Machina over Bluemix cloud services.