WP3: Dispatch optimization and machine learning

The aim of this work package is to provide software algorithms that can be used for optimization of plant operation by using forecasts of irradiation and electricity prices with associated uncertainty. For this purpose, an existing optimization approach is adapted to the new boundary conditions (other CSP configuration, accounting for forecast uncertainties). In parallel, machine learning algorithms are developed that allow parameterization of the plant performance models with up-to-date information derived from operational data. The work package is spilt into task 3.1 which defines the reference system set-ups, task 3.2 dealing with dispatch optimization and task 3.3 on machine learning. This work package is led by: DLR.

Objectives:

  • Develop and validate algorithms that provide an operation schedule for optimum dispatch operation to maximise plant revenue
  • Extend existing scheduling optimiser tools to handle various statistical uncertainties in inputs, e.g. from weather forecast and electricity market pricing
  • Extend existing scheduling optimiser tools to handle disturbances in the system operation, e.g. requests of grid support
  • Evaluate and implement machine learning concepts to improve plant performance mapping and integration into dispatch optimisation
  • Implement scheduling optimiser within demonstration control system
  • Develop all protocols and interfaces that allow implementation of optimiser within a full scale plant
  • Run the optimisation strategy and evaluate the business case on the demonstrator
  • Evaluate, track and control technical risks