Robust optimisation of electricity supply systems
With the switch to an energy provision that in its ultimate version is intended to be fed almost exclusively from renewable sources, the ‘uncertainties’ provided by these sources are posing considerable challenges. ‘Uncertainty’ in this context means that the output from renewable energies is dominated by factors that can be extremely difficult to predict with reasonable accuracy.
The wind, solar hours and heating requirements sometimes differ significantly from the forecasts. Technical and market uncertainties such as machine failures or price volatilities provide in their totality a very complex starting position for planning the power supply on a daily basis. Added to this is the fact that renewable energies often form a so-called virtual power plant comprising numerous, geographically widely distributed individual systems of various types and with sometimes low outputs, which require a correspondingly decentralised load management.
In order to reliably generate schedules that are designed to cover, for example, the electricity requirements for the next day as precisely as possible, so-called deterministic optimisation methods have been used until now. Thanks to stable forecasts and the availability of large centralised power plants, these are able to determine the absolute optimum for economic power plant operation. The described situation means, however, that this will soon no longer be possible in the near future, as even small changes to one of the currently dominant uncertainty factors could lead to large forecasting errors and thus to greatly different optimisation results. This particularly applies when several, possibly even correlating variables are situated in a wide band of uncertain values. Despite the circumstances described, it is therefore important to find mathematical methods that are able to respond sufficiently robustly, i.e. provide an optimal, stable result with a starting situation with widely varying values. As before, the provision of ancillary services such as primary and secondary control reserve and minute reserve shall be considered in the optimisation calculation. The goal of the research project considered here is to find, implement and test powerful mathematical methods that meet the aforementioned requirements and thus help to ensure a secure and stable power supply.
At the start of the project the BoFiT© optimising tool from ProCom GmbH was presented and tested in terms of its suitability as a platform for calculating the likewise determined application areas. The following schematic diagram shows the context of the energy industry objects modelled with BoFiT:
For the first application it was decided to investigate a cogeneration (CHP) plant in combination with wind and solar power infeeds, whereby the electricity price, heat requirement and of course the output from the wind energy and photovoltaic systems are to be included as uncertain variables. Other possible applications include minimising the physical grid losses, modelling reactive power in distribution grids and modelling the investment planning – each with their critical uncertainty factors.
Four phases to the project’s success
Phase A: Delineation of the problem areas
Phase B: Problems with only one type of uncertainty (one-dimensional problems)
Phase C: Problems with multiple types of uncertainty (multi-dimensional problems)
Phase D: Evaluation of the results obtained with real-time data
The project is currently in Phase C "Dealing with issues concerning multi-dimensional problems" and the start of Phase D "Evaluation".
The project partners and their tasks
RWTH Aachen University: As part of the development of innovative methods and concepts for further exploiting the potential provided by decentralised load management with the inclusion of ancillary services, RWTH Aachen University shall develop new kinds of robust optimisation models. For this purpose, suitable "uncertainty sets" for the uncertainty factors will be defined and the deterministic optimisation problem expanded accordingly. Key aspects of the project include the correct statistical description of the uncertainty of the observed variables and their mathematical description in the optimisation model as well as the investigation of correlated uncertainties and their effect on the overall model.
ProCom GmbH: The company shall develop prototypical robust optimisation methods for selected power plant fleets and for the load management. The intention is to primarily observe generating plants with uncertainty factors, including CHP plants, wind and PV systems. For validation purposes, the methods developed are being tested with selected customers using existing production systems.
10/2014 – 12/2016
52070 Aachen, Germany